Testkube Changelog

info

This is the changelog for the Testkube Control Plane, latest releases are on top.

Monthly releases add new functionality and overall improvements.
Patch releases focus on critical bugs and security issues.
Doc updates focus on major changes to the documentation

Release v2.15.0 (2026-10-29)

AI Governance with access, model and token limits, GitLab support and a new Test Bundle layout for AI Test Creation, an executions timeline, and HTTP Event Triggers.

Control Plane is available in Helm chart v2.338.0 and the Runner is available in Helm chart v2.15.0.

Upgrade notes

Self-hosted: give the worker service and the Git provider onboarding worker the same MCP encryption key as the API (ai.encryptionSecretRef), or GitHub PR analysis can't read the keys AI Agents sign in with. Read more.

AI Test Creation

GitLab repositories in Test Bundles: Import a GitLab repository into a Test Bundle, review a chat's changes and push them back to GitLab with a link to open a merge request, just as with GitHub. Connect GitLab with OAuth or an access token in the organization settings under Git, which now holds both GitHub and GitLab.
AI Agent Delegation in Test Bundles: Test Bundles can now delegate tasks to AI Agents, with access to MCP tools through delegated permissions. This enables a wide range of workflows, from bringing external context into a test bundle and splitting work across agents to keeping task management systems up to date. Admins can configure which AI Agents are available by default for new Test Bundles in AI Settings → Test Catalog, and customize them for individual bundles in Settings → AI Agents. Private AI Agents can't be added to a Test Bundle or to the defaults for new bundles, and making an AI Agent private removes it from the bundles that use it.
Start a new chat from any branch of the bundle's repository. The New chat dialog lets you write the first prompt and pick the model and the branch to start from, and pushes target that branch. Each chat runs its own files, and shows only its own executions.
New Test Bundle layout: the chat sits next to Files, Changes, and Past Executions tabs, and Settings in the bundle header holds General, Git repository, AI Agents, and Attached Context. Read more.
Review a chat's changes as a list of changed files with expandable diffs, inline or side by side, and push them from the Changes tab. Read more.
See an execution's artifacts, such as reports and screenshots, in the Artifacts tab next to Log Output.
Rename and delete chats, and copy a link to a chat. Opening a Test Bundle reopens the last chat you worked in. Read more.
A Test Workflow's Test Catalog tab lists the Test Bundles and chats that use it, and an execution started from a Test Bundle shows its bundle and chat.
Test Bundles created without a repository can place their tests in a folder of an existing repository when you connect one later.
Generated Playwright projects produce JUnit, HTML and trace reports, which appear in Artifacts.
Importing a repository from a chat now requires being a participant in that chat and having a Test Creation seat, and the import is recorded in the audit log.

AI Governance

Control who can use AI, with which models and how much, from Organization management → AI Governance, which brings Access, Models, Limits, Tracing and the Audit log for every Environment into one place.

AI Agent access: Make an AI Agent Public or Private, or limit it to Roles or Resource Groups. Members only see and use the AI Agents they have access to, and the Chats page lists only chats with those AI Agents. Read more.
Model access: Limit which models an AI Agent or a Member can use. Chats only offer the models a person may use, and a chat on any other model is refused with a clear message. Read more.
Token limits and pauses: Set token limits for the organization, an Environment, an AI Agent or a Member; the lowest limit that applies wins. Reaching a limit stops the chat mid-turn and refuses new chats and AI Triggers until it resets. Pause and resume AI for the whole organization. Read more.
Token usage: AI Usage shows tokens used per Environment, AI Agent, Member and model, and the usage ring shows tokens used against the limit as you chat. Read more.
AI tracing: Send a trace of each AI turn to AWS CloudWatch or an OTLP endpoint, and send a test trace to check the setup. Read more.
Audit log: MCP tool calls are recorded as MCP Tool Called, with the tool and the MCP Server. Filters are kept in the URL, so you can share a filtered view. Read more.
MCP sign-in per AI Agent: Each MCP Server an AI Agent uses has its own sign-in: a shared key, which can come from a Credential; each person signs in with OAuth; or none. MCP Servers now hold only their name, URL and OAuth client, and existing keys move to the AI Agents that use them. Under the chat input, each server's connection state is shown, and people are asked to connect when a server needs their own sign-in.
OAuth MCP Servers that don't support dynamic client registration accept a client ID and secret entered by an admin.
AI Triggers and GitHub PR analysis can't use an AI Agent whose MCP Server needs each person to sign in; the picker greys it out.

AI

Test Authoring now uses the model and thinking level selected for the chat for its work in the Sandbox.
Claude models are supported natively, with thinking levels and structured output, directly or through a gateway such as OpenRouter. Read more.
Pick the model and its thinking level from one searchable list.
The MCP Server tab adds one-click install for Cursor and VS Code and setup steps for Codex and Copilot CLI. Read more.
Analyze a finished execution in a local AI tool, such as Claude Code, Codex or Cursor: Open in in the execution header opens the tool with a prompt about that run, which you review before sending.
The AI Assistant and MCP tools can list, read, create and edit Insights boards.
New AI chats are private by default. You can make a chat public when you start it.
If the selected model is unavailable, a chat falls back to the default model instead of failing.

Executions

Timeline view: Switch the Executions page to Timeline to see each Workflow's Executions as bars over time, which Executions were running at any moment, and the CPU or memory they used. It uses the same filters and Custom Views as the list. Read more.
Logs: A faster log viewer for large logs. Show the logs of every step at once with Show all logs, and copy only the highlighted lines with Copy highlighted lines. Read more.

Fixes & Improvements

A Test Workflow's definition refreshes after an AI Agent updates it, without discarding your unsaved edits.
Runners connected during a session now appear in the Run tests runner list.
Test Authoring no longer starts a run when no Runner can take it, and says so in the chat.
Repository imports explain when the GitHub App cannot access the repository, with a link to its repository access settings.
Test Authoring Sandboxes reach the Control Plane API when serveHTTPS is enabled.
MCP clients no longer get intermittent sign-in errors after their access token refreshes.
MCP Servers that answer notifications with 204 No Content now connect.
Runners with only the GitOps capability are no longer marked as disconnected in the Control Plane.
Prometheus metrics have lower cardinality. Per-step execution metrics stay off unless testkube-cloud-api.api.features.workflowStepMetricsEnabled is enabled (TK_WORKFLOW_STEP_METRICS_ENABLED), and HTTP handler metrics use the route template instead of a separate series for each endpoint.
CVEs clean up: resolved all CVEs for which fixes are available.
Upgrade NATS and Dex: now NATS is in version 2.14.9 and Dex in version 2.46.1.
Removal of emSign root certificate: removed eMudhra intermediate CAs and other risky certificates.

Runner & CLI

Features

HTTP Event Triggers: A TestTrigger can poll an HTTP endpoint and run a Test Workflow when the response changes: a JSON field, a response header such as ETag, or the whole body. Read more.
MongoDB to PostgreSQL migration for the standalone agent: The convert job copies past Test Workflow executions and execution numbers from MongoDB to PostgreSQL, and resumes where it stopped if interrupted. Read more.

Improvements

testkube upgrade --reuse-values keeps the values set at install time instead of resetting them to the chart defaults. Read more.
testkube diagnostics agent checks that the Runner's API server pods are ready and names the reason when one is not. Use --output json to read the results from scripts.
File flags of the CLI accept - to read from standard input, for example cat payload.json | testkube create webhook --payload-template -. This covers webhook and webhook template payloads, create testworkflow and create testworkflowtemplate with --file, and agent install --global-template-path.
--verbose shows the telemetry payload the CLI sends.
A burst of scheduled executions, for example many cron triggers in the same minute, no longer stops a standalone agent from starting executions.

Fixes

Restored caches keep the modification times of their files, so tools such as the Go build cache can trim old entries and the cache no longer grows on every save.
With more than one API server replica, an execution started twice is no longer declined while it runs.
The init process closes connections on its control port that send anything other than a pause or resume command, so a TLS probe or a port scanner can no longer pause a step.
Execution notification streams stop when the client disconnects, and the CLI's watch commands use them.

Patch Release v2.14.2 (2026-10-14)

Control Plane is available in Helm chart v2.337.2 and the Agent is available in Helm chart v2.14.2.

Upgrade notes

Personal organizations are now disabled by default, so new users no longer get a personal organization on first sign-up. To keep the previous behavior, set testkube-cloud-api.api.features.disablePersonalOrgs: false in your Helm values.

Security

CVE cleanup: Upgraded bundled third-party components to resolve known CVEs: MongoDB 8.3.11, PostgreSQL chart 18.6, SeaweedFS chart 4.44 and kubectl 1.37.1.

Fixes

Test Workflow execution output data: Fixed an issue with saving output within child workflows.

Patch Release v2.14.1 (2026-10-05)

Control Plane is available in Helm chart v2.337.1 and the Agent is available in Helm chart v2.14.1.

Fixes

Hardened MCP Authorization logic.
MCP usage with API tokens is now correctly billed to the organization instead of blocked.
CVE cleanup: Resolved all CVEs for which fixes were available.

Release v2.14.0 (2026-10-01)

GitHub support and multiple chats for AI Test Creation, clearer execution failure diagnostics, and new workflow capabilities for dependency caching and data sharing.

Control Plane is available in Helm chart v2.337.0 and the Runner is available in Helm chart v2.14.0.

Upgrade notes

The Control Plane API readiness probe now uses /health/readiness, which checks dependencies. AWS S3 installations must allow s3:ListAllMyBuckets on the API's IAM role. See S3 with IAM Authentication.
PostgreSQL installations using separate migration and runtime roles should verify default privileges and access to newly created tables. See the Post-Upgrade Runbook.

Workflow orchestration compatibility change

Starting with Testkube 2.14.0, separate entries in the same execute.workflows block must have distinct execution references. An entry uses its as alias when provided; otherwise, it uses the Workflow name.

Existing Workflows that list the same Workflow more than once without distinct references will fail with duplicated execution reference. This validation applies even when the Workflow does not use execution() or share data between executions.

Before upgrading from 2.13.x, give repeated entries distinct aliases:

execute:
  workflows:
    - name: my-workflow
      as: my-workflow-a
    - name: my-workflow
      as: my-workflow-b

Aliases must also be distinct from the references of other entries in that block. A single entry expanded through count, matrix, or shards remains supported without assigning an alias to each instance.

See Advanced Workflow Orchestration.

AI Test Creation

Work with a GitHub repository inside a Test Bundle. Import the repository, review file changes, and push the chat's commits back to GitHub. Read more.
Test Bundles now support multiple chats, each with its own changes. Explore different approaches or work on separate test updates within the same bundle. Read more.
Add shared context to a Test Bundle to give every chat the same project guidance and instructions. Read more.
Review changes with file diffs and highlights showing what has changed since your last push. Read more.

AI Test Creation is now generally available. See Create your first test with AI to get started.

Workflows

Cache dependency directories between executions with a step-level cache block. Use file-content hashes for keys, fallback restore keys, and workflow or environment scope. A cache miss or transfer failure leaves the step free to install its dependencies normally. Read more.
Share output values and artifacts between related Workflow executions. Suites can read child outputs, read artifact contents in expressions, and fetch artifact files into another execution. Read more.
Build workflows that rerun only previously failed test cases. Execution lineage identifies the original run, the previous attempt and the attempt number; workflows can use the previous report to select what to rerun. Read more.
Configure multiple Kubernetes events in one TestTrigger. Existing single-event trigger definitions remain supported. Read more.

Execution Diagnostics & Insights

Execution status now explains whether a run ended because of a test failure, infrastructure failure, configuration error or user cancellation. Cause details are available in the dashboard, CLI and MCP tools, including clearer initialization, timeout and retry information.
Filter executions and segment Insights by failure cause. Workflow health and charts use consistent failure counting, with user cancellations excluded from failures.
Subscribe webhooks to specific failure categories. GitHub checks and pull-request comments now include the cause of failed workflow executions, making it easier to distinguish test failures from infrastructure or configuration problems.
Link directly to an individual JUnit test case in a report.

Members & Audit Logs

Expanded member management page with new filtering, bulk operations, and clearer member details.
Search audit logs by name or ID to quickly identify all actions related to a resource.

AI

Choose the thinking level for models that support it. The dashboard AI panel can now be resized and collapsed, and AI chats and Test Bundles receive shorter, context-aware titles.
Where AI usage controls are enabled, administrators can review session and MCP usage, manage Test Creation seats, and see clearer warnings as usage approaches a plan limit. Availability and enforcement depend on the installation and plan configuration.

Operations & Improvements

Replaced all references to Agent with Runner. This includes pages and titles within the UI as well as flags and commands from the CLI. The agent commands/pages have been marked as deprecated and will be removed in the next major version.
Added support for the execution capability, replacing the deprecated runner capability.
On-prem Sandbox CPU and memory requests and limits are configurable. An optional prewarmed Sandbox pool can reduce startup work for new authoring sessions.
Improved completion and cancellation handling, including workflows whose test containers have finished while a sidecar is still running, and stale queued executions that should retain their history when aborted.
Improved resilience of artifact transfers, webhook delivery and cleanup during transient failures. Invalid GitOps resources no longer cause continuous retries.
Added new options --org-name and --env-name to login
Support --client=direct against a control plane.

Patch Release v2.13.3 (2026-09-22)

Agent only. The Agent is available in Helm chart v2.13.3; the Control Plane stays on v2.336.2.

Improvements

Internal improvements to the On-Prem trial experience.

Patch Release v2.13.2 (2026-09-22)

Control Plane only. The Control Plane is available in Helm chart v2.336.2.

Fixes

Fixed workflow template Definition view failing to load in certain conditions.
Template updates now answer with the updated template on source-of-truth environments.

Release v2.13.1 (2026-08-27)

Advanced GitHub-native integration, test case-level Health metrics and Insights, plus many smaller improvements and fixes.

Control Plane is available in Helm chart v2.336.1 and the Agent is available in Helm chart v2.13.1.

Advanced GitHub Integration

Connect a GitHub repository to an environment. Testkube can generate Test Workflows from the repo, or use ones you already have, then run them on pull requests, pushes, tags, and releases.
Results land on the pull request. Testkube comments pass/fail with the workflows that ran, can add an AI summary of failures, and can fail a GitHub Check so branch protection blocks the merge.
Read more about using the advance GitHub Integration here.

AI

Direct links to chat messages. Every message in an AI chat now has its own link, so you can point a colleague at a single answer instead of the whole conversation.
Agents link their sources. When an agent inspects an execution or a workflow to answer you, its reply now links straight to it, so you can check the run behind the explanation.

Insights & Analytics

Test case health: Insights now tracks the health of individual test cases. This can be used to identify flaky tests or view health of a test case over time.
Updated coloring of heatmap to more closely align with chosen metrics/values.
Fixed an issue with JUnit reports not being correctly identified and parsed

Operations & Security

Testkube AI authorization: AI, MCP and agent endpoints now share a single set of authorization guards, with caller identity resolved through the Control Plane. Permission enforcement is consistent across every AI surface and easier to audit.
Helm: the AI service internal secrets are now configurable from the umbrella chart.

Improvements

Added an option to rerun an execution using the latest workflow definition (but reusing the same execution parameters)
Added two new options when using spec.content.git: verbosity to control log levels and retry to control retry behavior - read more

Other

Added support for explicit GitOps agent ownership - read more
Fixed an issue where errors were not displayed properly when creating a new workflow using the workflow wizard.
Testkube's own MCP server is now published in the public MCP Registry with its hosted endpoint, so clients looking for remote MCP servers can find and connect to it.
Webhooks ignore silent executions. A failed silent run was still used as the previous workflow state, so the next passing run could fire become-testworkflow-up. Silent executions are now excluded from webhook status and state transitions.

Patch Release v2.12.2 (2026-08-18)

Control Plane is available in Helm chart v2.335.2 and the Agent is available in Helm chart v2.12.2.

Improvements

Hide AI Panel: Hide the AI panel on the home page when the AI feature is disabled for the organization.
Prometheus metrics:
- TestWorkflow execution metrics now include workflow_namespace.
- Webhook CRUD operations now emit Prometheus counters for create, update, and delete events.
- GitOps/sync-based TestWorkflow and TestTrigger CRUD operations now appear in Prometheus metrics, including error outcomes.
- A new testkube_testtrigger_events_total metric tracks workflow executions started by a TestTrigger, labeled by trigger name.
- Optional metadata metrics can now export custom TestWorkflow labels through testkube_testworkflow_custom_label_info. For more details, see the Control Plane Metrics documentation.
Patch 3rd party components: patching images for MongoDB and PostgreSQL.

Fixes

PostgreSQL migration: migration script was fixed due duplicated file name which generates an error of missing column.
CVE cleanup: Resolved all CVEs for which fixes were available.

Patch Release v2.12.1 (2026-08-04)

Control Plane is available in Helm chart v2.335.1.

Improvements

Testkube CLI: The new testkube pro export usage command simplifies exporting usage data from air-gapped Enterprise installations. It automatically detects the Control Plane configuration and creates a ZIP file that customers can review and share with Testkube when requested.

Fixes

Testkube AI security: Strengthened authorization across Testkube AI, MCP servers, and AI agents to ensure access consistently reflects each user’s permissions.
PostgreSQL support: Test Workflows without executions are now visible to all users with the appropriate permissions, and pagination now displays the correct number of pages.

Agent

Added support for matchExpressions in TestTrigger workflow selectors, enabling more flexible label-based TestWorkflow matching.

Release v2.12.0 (2026-07-23)

Private AI chats, deeper Insights and Analytics, plus major improvements across security, observability, CLI workflows, and runner operations.

This is a major release with lots of new features and improvements!

Control Plane is available in Helm chart v2.335.0 and the Agent is available in Helm chart v2.12.0.

AI

Private AI chats. Start a private chat to keep the conversation visible only to you, then make it public when you're ready to share it with your team. Making a chat public is permanent.
AI chats in workflow and execution context. Workflow details now include an AI Chats tab, while the existing execution tab now shows any chat where an agent inspected that workflow or execution, not only chats started by triggers, making related analysis easier to find and continue.

Insights & Analytics

Added support for execution drilldown on all metrics
Added support for filtering or segmenting by execution tags in time series analysis
Extended available segments for dynamic and custom metrics

Operations & Security

Improvements

Registration token rotation: from the Testkube CLI, it is now possible to rotate the registration token that agents use to self-register with the Control Plane. Once updated, the token must be updated in all locations where it is used: values.yaml files and/or secret stores.
Testkube CLI supports --limit for queries: the CLI now allows the --limit parameter for the Test Workflow get method to return more than 100 objects.
CVEs resolution: all vulnerabilities (CVEs) reported in previous versions of the images have been resolved, for both the Control Plane and Agent artifact groups.

Other

Added ability to filter executions based on actor and execution time
Improved error messages when editing workflow and other definitions via YAML
Added extra details to logs pertaining to workflow, execution, and runner metadata
Replaced outdated Grafana dashboard metrics with their current corresponding ones
Added an option to expose Prometheus metrics for runners
Improved runner logging to reduce false error messages

Patch Release v2.11.3 (2026-07-17)

Control Plane is available in Helm chart v2.334.3.

Fixes

Fixed an issue with mongo to postgres migration
Fixed an issue where the AI service would be unable to communicate with the Control Plane

Patch Release v2.11.1 (2026-07-02)

Control Plane is available in Helm chart v2.334.1.

Security

MCP Server security hardening. Strengthened authorization and permission enforcement across the MCP endpoint and AI provider APIs, so access consistently respects organization, environment, and role-based permissions.

Monthly Release v2.11.0 (2026-06-25)

Control Plane is available in Helm chart v2.334.0. and the Agent is available in Helm chart v2.11.0.

This is a major release with lots of new features and improvements!

AI Updates

AI in Testkube is now easier to configure, faster, and safer to run.

Create AI agents from goals: describe what you want the agent to do, and Testkube generates the agent name, description, instructions, and recommended tools.
Configure a default model per AI agent, so each agent can use the model best suited for its work.
Choose the model when replying in an AI chat.
Cancel running AI chats from the UI.
AI chat titles now appear immediately and continue to improve in the background when needed.
AI chat responses are faster thanks to improved streaming behavior and reduced per-request overhead.
AI triggers are safer: empty selectors no longer match every execution, wildcard matching is explicit, and trigger sessions can be rate-limited and manually stopped.
MCP OAuth errors are clearer and include actionable guidance when the MCP endpoint is disabled.
Testkube MCP tool descriptions were trimmed to reduce prompt size and improve AI response performance.
Self-hosted AI Service configuration is simpler: the service now discovers dashboard configuration from Control Plane public config and no longer requires several OIDC-related environment variables.

Test & Git Triggers

Testkube can now observe a Git repository and trigger executions on repository events. This covers the full lifecycle from simple branch pushes to GitHub pull-request automation, with fine-grained filtering and rich Git context exposed to the triggered execution (read more)
Test Triggers can now watch custom resources and fire on specific field changes, not just built-in Kubernetes events (read more). Point a trigger at any CRD, such as Argo Rollouts, cert-manager, or your own, and add match conditions so it only runs when a field you care about changes, for example when a Rollout's .status.phase becomes Healthy. The dashboard discovers the resources each listener agent can watch, autocompletes their fields from the live schema, and lets you pin the trigger to the listener that should evaluate it

Onboarding

Streamlined the migration onboarding experience, making it easier for Testkube Open Source users to get started with Testkube Cloud
Trial users of our Testkube Cloud control plane now get an automatically provisioned, cloud hosted runner set up for them so they can start running test workflows straight away. Users can still add their own agents as well, of course.

Insights Updates

Enhanced the Insights Board user experience with support for pinned boards, shared/privated boards, and search results by name - (read more)
Added support for label and workflow filters searching for a substring (contains)
Added support for identity fields searching for a substring (contains)
Added a default Insights Board at the Workflow level with a default set of charts - (read more )
Added eight chart types in Insights for better analysis of execution metrics (read more)
Granular Metrics
- Native parsing of k6, Artillery, and JMeter metrics into Insights - (read more)
- Canonicalized metric names: k6, Artillery, and JMeter now share normalized latency_p95_ms, throughput_rps, and errors_rate alongside the tool-specific names, so you can compare runs across tools in one chart and filter by source - (read more)
- Added support for Influx Line Protocol for creation of custom metric series - read more

Operations & Security

Testkube can now authenticate to GitHub through a single centralized Testkube GitHub App. One Testkube organization can connect to multiple GitHub organizations and access repositories across all of them (read more)
Improved the Usage & Billing experience with transparent consumption tracking for all plan levels, updated every hour
Testkube's SCIM integration now supports the RFC 7643 complex format for roles and entitlements, improving provisioning with Microsoft Entra ID (Azure AD). Identity providers send these multi-valued attributes differently: Okta and most providers use simple string arrays, while Entra expects each value wrapped in an object. Testkube now handles both, and a per-organization setting lets you choose the format that matches your provider.
Added a new maxRole ceiling that lets you lock a member's effective permissions to read-only regardless of the roles they hold on individual environments or teams (read more)
Helm Charts Improvements
- Added Gateway API support for Testkube Enterprise, including configuration for Gateway, HTTPRoute, HTTP-to-HTTPS redirect, BackendTLSPolicy, and certificate configuration options, and with migration guidance for moving from Ingress without downtime. The Gateway API integration works with any compatible Gateway controller, with Traefik provided as the primary example, although controller-specific behavior and configuration may vary. NGINX Ingress remains supported for existing installations, but Gateway API is recommended for new deployments.

Runner Agent

Runner lease coordination configuration: New environment variables on the agent to control its Kubernetes lease coordination:
- LEADER_ELECTION_DISABLED -> disables lease coordination for explicitly single-replica runner deployments . Default: false
- LEASE_CHECK_INTERVAL -> how often the agent checks/renews the lease (e.g. 30s, 1m). Default: 5s

Patch Release v2.10.3 (2026-06-09)

Control Plane is available in Helm chart v2.333.3.

Incremental adjustments with no impact on enterprise/on-premise deployments.

Patch Release v2.10.2 (2026-06-07)

Control Plane is available in Helm chart v2.333.2 and the Agent is available in Helm chart v2.10.1.

Improvements

License calculations: improve and fix calculations for fixed and floating agents.
MCP server support custom certificates: custom CA is now configured as part of AI Service to trust in self-signed certificates.
Agent labels and mode: manage and update agent labels and mode from helm chart values.
TestWorkflow execution queue metrics: new metrics to monitor TestWorkflow execution queue which includes metadata information useful to know what TestWorkflow is in the list.

Fixes

User limits in Dashboard: members page correctly shows the user limit.
TestTrigger templates: support JSON-style field names (e.g. .metadata.labels) for built-in resources, matching resourceRef.
Cron job target configuration: cron job configuration in TestWorkflow Template allow target property to schedule executions in specific agents.
TestWorkflow multiple target conditions: moving from OR to AND for all target conditions defined in a TestWorkflow execution.

Monthly Release v2.10.0 (2026-06-01)

Control Plane is available in Helm chart v2.333.0 and the Agent is available in Helm chart v2.10.0.

What's new!

Granular Test Insights

In v2.10.0, we've expanded the underlying metrics data model for Test Insights so you can do a deep dive into your E2E/functional and performance test results over time.

Automatically process JUnit, JMeter, k6, and Artillery reports, extracting relevant metrics and exposing them in the Insights.
Filter and segment results by test case, test suite, endpoint, and other relevant dimensions.
Explore extended metrics with improved metric grouping in the Insights time-series picker.
Improved ordering of stacked segments to be consistent across columns.

AI Updates

AI chats now surface clearer progress while work is in flight, and agents can request credentials securely when they need access to secrets — without putting values in the thread. Together, these changes make longer AI tasks easier to trust and follow from start to finish.

Testkube MCP Server now supports OAuth as the preferred authentication method.
Much improved configuration for custom models in on-Prem/Air-gapped deployments.
New "AI Analyze" actions in the Execution and Workflow views that default to use the corresponding default AI Agents included with Testkube
Display of Token Usage for Chats in the Chats Overview and Details panel.
The Open-Source Execution Viewer now allows for AI-generated summaries of test results so readers can quickly understand failures and what to investigate next — alongside the JUnit report browsing added in recent patches.

Improvements

Clearer status and activity messages while the agent is working.
Secure credential requests in chat without exposing secret values in the conversation.
More transparent chat behavior during longer-running tasks.

Fixes

Fixed several AI chat UX issues, including cases where chat titles or message input behavior could become inconsistent.

Control Plane

On-Prem Customers on the Enterprise plan can now manage multiple organizations from one place: create organizations, review usage across them, and get warnings before hitting plan limits.

GitOps-managed workflows how show a warning in the Dashboard for that edits may be overwritten by sync.

Fixes

Fixed an issue where switching organizations would occasionally show stale data.
Fixed an issue where the SCIM setup URL could be generated incorrectly.
Fixed the dashboard silent execution toggle so silent runs behave as expected.
Fixed cases where canceled parallel workers could still appear to be running in the UI.
Improved reliability of the executions view for large environments.

Agent

This release improves agent reliability for webhooks and execution-related flows, and makes execution lists faster and more dependable in large, high-churn environments.

Fixes

Fixed issues where webhooks or execution-related flows could behave incorrectly in some environments.
Improved execution-list performance and reliability for large-scale environments.

Open Source Updates

Templated Security Contexts

TestWorkflowTemplates can now use integer config parameters in security context fields such as runAsUser, runAsGroup, and fsGroup. This lets reusable templates render valid Kubernetes pod and container security contexts while keeping those values configurable.

AI Analysis in the Public Execution Viewer

Sharing a workflow execution publicly is useful for debugging handoffs, but raw logs and reports still take time to interpret. The public execution viewer now includes AI-generated summaries of test results so readers can quickly understand failures and what to investigate next — alongside the JUnit report browsing added in recent patches.

Patch Release v2.9.7 (2026-05-29)

Control Plane is available in Helm chart v2.332.7.

Improvements

PostgreSQL performance improvements: Improved PostgreSQL query performance.

Patch Release v2.9.6 (2026-05-28)

Control Plane is available in Helm chart v2.332.6.

Improvements

PostgreSQL performance improvements: Improved PostgreSQL query performance.

Patch Release v2.9.5 (2026-05-15)

Control Plane is available in Helm chart v2.332.5 and the Agent is available in Helm chart v2.9.5.

Improvements

MongoDB to PostgreSQL Migration: Added skip_tables parameter to the convert utility for Mongo to Postgres migration.

Fixes

Fixed an issue where executions lists would be truncated to a single page.
SCIM Support: Fixed SCIM support for Postgres based Control Plane.

Patch Release v2.9.4 (2026-05-07)

Control Plane is available in Helm chart v2.332.4 and the Agent is available in Helm chart v2.9.3.

Features

JUnit Report Tab in Public Execution Viewer: The OSS execution viewer now includes a JUnit report tab, allowing users to browse test results directly in the public viewer.

Improvements

PostgreSQL: Execution query performance has been improved through denormalization migrations and updated insights queries.

Fixes

Fixed an issue where agent capabilities were not correctly queried, which could lead to incorrect agent state in the UI.

Patch Release v2.9.3 (2026-05-05)

This is an Agent only release. the Agent is available in Helm chart v2.9.3.

Improvements

PostgreSQL Support: The execution time for some queries has been reduced when using the PostgreSQL database.
Skip Kubernetes TLS verification: The agent API can now handle communication to collect logs from running pods when the Kubernetes API uses an invalid TLS certificate.

Patch Release v2.9.2 (2026-04-30)

Control Plane is available in Helm chart v2.332.2 and the Agent is available in Helm chart v2.9.2.

Chart Fixes

Missing IMPORTS_BUCKET variable: Added the missing IMPORTS_BUCKET environment variable for both cloud-api and worker services. By default, it now falls back to OUTPUTS_BUCKET unless a different imports bucket is explicitly configured. This variable is used for importing OSS data into Testkube Cloud; because it was missing, upgrades of the charts to v2.9.1 were failing.

Features

Disable test log persistence: Added a none option for logs.storage to disable test log persistence entirely for environments that don't need to retain log data.
S3 Virtual-Hosted-Style URLs: Added support for S3 virtual-hosted-style URLs (bucket.s3.region.amazonaws.com) alongside the existing path-style addressing.

Fixes

Container security context on init containers: Fixed an issue where ContainerSecurityContext could not be applied to init containers in TestWorkflow pods.
Floating runner parallel scheduling: Fixed an issue where an already-active floating runner could be blocked from taking additional parallel executions when the floating runner concurrency limit was set to 1.
Quiet workflow log streams: Fixed an issue where workflow log streams could disconnect during quiet periods even though the workflow was still running.
TestWorkflowExecution controller status generation: Fixed Status.Generation never being written by the testworkflowexecution controller, which caused all TWE CRDs to re-trigger on every operator restart.

OSS Execution Viewer

Validate execution before prompting: testkube view now validates the execution ID (or name) before asking for upload permission, so users see a clear "execution not found" error instead of being prompted for non-existent executions.
Log panel scroll: Fixed an issue where the execution viewer expanded log panels to the full log height, breaking the internal scroll. Logs now scroll within their fixed container.
Avoid duplicate uploads: testkube view <id> now checks whether the execution has already been shared and reuses the existing public link instead of re-uploading the same execution.

Monthly Release v2.9.1 (2026-04-23)

Control Plane is available in Helm chart v2.332.1 and the Agent is available in Helm chart v2.9.1.

What's new!

View Open Source Execution Results

Open Source users can now view execution details (step tree, logs, and artifacts) from their TestWorkflow executions directly in their browser, in a lightweight, read-only page hosted by Testkube. No login or Cloud account required.

To try it, update to the latest CLI and run:

testkube view <execution-id-or-name>

and it will open the generated link in your browser.

AI Updates

Interactive Clarification in AI Chats

AI agents can now ask structured follow-up questions when your request needs more context. Instead of making assumptions, the agent presents an interactive form to gather the details it needs — whether you're creating a workflow, debugging a failure, or exploring execution data — then acts on your answers.

Reconnect to In-Progress AI Sessions

Navigate away from an active AI chat and return without losing context. Sessions continue in the background, and all prior events are shown when you reconnect. Multiple users can observe the same session simultaneously.

Improvements

Users can manually rename chat titles from the sidebar.
Chat overlay persists across page refreshes.
AI agents can now analyze logs from parallel workflow workers — use grep, line ranges, and step filtering on individual worker output, not just the main execution log.
AI agents can now filter executions server-side by tags, time ranges, and label selectors, enabling precise queries like "all executions longer than 30s without a suite label" without downloading and scanning every result.
MCP registry entries now indicate the expected authentication method when browsing available servers.

Fixes

Fixed artifact file retrieval when reading files in subdirectories through the MCP server.
Fixed date and time filtering for execution queries using natural-language time ranges.
Fixed JSON path queries that could miss matching workflows.
Deleting an AI agent now removes associated AI triggers automatically.

Triggers Updates

Custom Resource Triggers

TestTriggers can now fire on arbitrary Kubernetes Custom Resources, not just the built-in set (Pods, Deployments, etc.). Any CRD installed in the cluster — Argo Rollouts, Flux, Crossplane, your own operator's types — can be named as the trigger's resource by its Group, Version, and Kind.

Field-Level Trigger Matching

TestTriggers now support spec.match to fire only when specific fields on the watched resource satisfy a condition. Instead of reacting to every create/update event, a trigger can watch for a precise state transition — e.g. status.phase == Healthy or status.currentStepIndex == 3 — and run the associated Workflow only then.

Example: Argo Rollouts

Combining the two, you can tie test execution directly into a progressive delivery pipeline. A single Rollout resource moves through phases as canary steps advance; with field-level matching you can kick off the right tests at the right step without polling or external glue:

Smoke tests after each canary step — match on status.phase == Paused to run quick health checks while the Rollout is waiting at a weight increment.
Full regression on completion — match on status.phase == Healthy and status.currentPodHash changing, so tests run exactly once per successful rollout and not on unrelated status updates.
Rollback validation — match on status.phase == Degraded to trigger a diagnostic workflow the moment Argo aborts a rollout.

Open Source Updates

`testkube pro connect`

Connects a Testkube Open Source installation to a Testkube Control Plane. Use this command when you want to:

Evaluate Testkube Pro functionality and capabilities with your existing TestWorkflows and results.
Migrate your existing Testkube resources to a commercial Testkube instance (cloud or on-prem).

Before switching, the CLI automatically exports all TestWorkflow execution data (metadata, logs, and sequence numbers) from the local agent and uploads it to the Testkube Control Plane, ensuring seamless data migration. It then installs a Testkube Runner Agent, scales down the now-unnecessary OSS services (API server, MinIO, NATS, and the database), and updates the CLI context to Pro mode. Users can skip the export with --skip-export or limit it to recent data with --since.

Features:

Automatic data migration: Execution data is exported as a .tar.gz archive and imported to the Control Plane in a single flow — no manual steps required.
Graceful error handling: If the export or import fails (e.g. archive too large), users are warned and can continue connecting; the archive is preserved for manual retry.
Database-aware teardown: The CLI auto-detects whether MongoDB or PostgreSQL is deployed and only scales down the active database.
CRD conflict prevention: The runner chart is installed with gitops.installCRD: false to avoid duplicate CRD ownership with the existing OSS chart.

`testkube pro disconnect`

Reverts a Testkube Pro installation back to OSS standalone mode. The command uninstalls the Testkube Runner Agent chart, deletes the agent record from the Control Plane, scales the original OSS services (API server, MinIO, NATS, and the database) back up, and resets the CLI context to kubeconfig mode. All historical data remains intact in the original cluster database.

Features:

Full automated rollback: The runner Helm release and Control Plane agent record are cleaned up automatically.
Database-aware restore: Scales up only the database type (MongoDB or PostgreSQL) that was originally deployed; falls back to attempting both for clusters connected before this feature was added.
Configurable replicas: Users can control how many replicas to restore for MinIO (--minio-replicas), MongoDB (--mongo-replicas), and PostgreSQL (--postgres-replicas).

note

disconnect reverts the Testkube Agent back to the standalone state it was in before it was connected to a Testkube Control Plane. Any resources or results that were created in Testkube while the agent was connected to the Control Plane will be lost.

`testkube marketplaces <command>`

Added CLI commands to browse and install TestWorkflows from the Testkube Marketplace - Read more

Control Plane

Improvements

Following live TestWorkflow logs is now more reliable for customers using the CLI or integrations. Reconnects resume from the right place instead of hanging, replaying the wrong data, or getting stuck when the Kubernetes log stream goes silent unexpectedly.
The JUnit Report tab now automatically opens the first failed test and also allows for filtering on skipped tests - Read more.

Images

Updated image registry references in Helm values: container images are now sourced from DockerHub instead of Google Artifact Registry, including dependencies such as MongoDB, MinIO, PostgreSQL, and kubectl. Helm charts will continue to be published and stored in Google Artifact Registry.

PostgreSQL

Added support for supplying the PostgreSQL DSN from Kubernetes secrets for all Testkube services.
Added support for running PostgreSQL with CloudNativePG, alongside the existing bundled PostgreSQL deployment to preserve backward compatibility. This includes support for installing the CloudNativePG operator and provisioning an operator-managed PostgreSQL cluster. Read more.
Added migration command to migrate MongoDB to PostgreSQL - Read more

Solve vulnerabilities

The following CVEs were solved by upgrading libraries and packages in the Testkube images: CVE-2026-34986 and CVE-2025-27144.

Upgrade 3rd party images and helm charts

MongoDB to 8.2.7.
PostgreSQL to 18.3.
KubeCTL to 1.35.4.

Agent

Allow Multiple GitOps Agents in the same namespace

Fixed the naming of the GitOps Agent Kubernetes resource to allow multiple instances of it in the same namespace to sync Testkube Resources to multiple Testkube environments.

PostgreSQL

Added an option to source the PostgreSQL DSN from a Kubernetes secret.
Added support for running PostgreSQL with CloudNativePG while keeping the existing bundled PostgreSQL deployment for backward compatibility. This includes support for installing the CloudNativePG operator and provisioning an operator-managed PostgreSQL cluster. Read more: PostgreSQL

Upgrade 3rd party images and helm charts

MongoDB to 8.2.7.
PostgreSQL to 18.3.

Fixes

TestWorkflows are now easier to run on OpenShift and similar restricted clusters. You can opt out of automatic pod fsGroup injection when your platform expects security policies to assign that value.

Patch Release v2.8.3 (2026-04-07)

Control Plane is available in Helm chart v2.331.3 and the Agent is available in Helm chart v2.8.3.

Control Plane v2.331.3

Improvement & Fixes

Licensing Limits logic: fix how is counting floating agents across environment.
UX in Create Trigger: Remove autocomplete from create new trigger form.
AI Assistant PostgresDB with SSL: Improved SSL configuration for integrating with PostgresDB instance.

Agent v2.8.3

Improvement & Fixes

Resilient git clone with basic auth: configure git credential store to survive HTTP redirects during clone.

Patch Release v2.8.2 (2026-03-30)

Control Plane is available in Helm chart v2.331.2 and the Agent is available in Helm chart v2.8.2.

Control Plane v2.331.2

Improvements and fixes

New! Testkube on-prem installations no longer require testkube-agent. New environments start with the Control Plane owning Testkube Resource management by default.
testkube init demo no longer installs the agent bundle.
SCIM Server lock contention: write lock contention now returns 409 instead of 429. This is the correct status per RFC 7644 Table 9 for POST/PUT/PATCH/DELETE conflicts. The library passes it through, and IdPs treat it as retriable.
Follow long Test Workflow executions: include a keepalive signal to avoid timeouts while following long Test Workflow executions from CLI and integration platforms.
SSO Authentication: fix SSO login failure caused by sending static "sso" as connector_id.

Agent v2.8.2

Improvements and fixes

Installation OSS Agent: remove operator and param as it's already deprecated, replacing by Testkube CRDs only.
testkube-agent is deprecated in the testkube-enterprise chart.
Testkube Agent Global Template: manage inline and named Global Templates from Testkube Agent Helm chart properties. When upgrading the testkube-runner Helm chart from 2.7.x, inline global templates must now set globalTemplate.inline: true. Without inline: true, the chart treats the global template as a named external template and requires globalTemplate.name.
Testkube Agent Name: allow override agent name from Testkube Agent helm chart property runner.name.
Follow long Test Workflow executions: improve the behavior when the underlying execution is healthy but the live log stream is interrupted by the network path between the CLI runner and the Control Plane.

Patch Release v2.8.1 (2026-03-30)

Control Plane is available in Helm chart v2.331.1 and the Agent is available in Helm chart v2.8.1.

Control Plane v2.331.1

Improvements and fixes

Added MongoDB jobs to upgrade from version 8.0.15 to 8.2.5. Read more
SuperAgents can no longer be removed before migration.

Agent v2.8.1

Improvements and fixes

Added MongoDB jobs to upgrade from version 8.0.13 to 8.2.5. Read more

Monthly Release v2.8.0 (2026-03-26)

This release is available in Helm chart v2.331.0. Control Plane and Agent versions are now unified at v2.8.0. (Check out Agent release notes).

What's new!

You can now sign in to our SaaS using your Google account. This adds another convenient login option alongside our existing authentication providers.

Workflows Updates

Vault Integration

You can now use HashiCorp Vault (KV v2) as a backend in Credentials API. Workflow secrets are read from Vault at execution time instead of being stored in Testkube's database. Useful for teams that already manage secrets in Vault and do not want to duplicate them. As of now, this is available only for on-prem installations.

Supports Kubernetes auth, static token auth, and Vault Enterprise namespaces.

Resource Group Credentials

Credentials can now be scoped to Resource Groups, sitting between Environment and Workflow in the precedence hierarchy.

Workflow Marketplace

The Dashboard now includes a workflow catalog where you can browse pre-built TestWorkflows, configure parameters, preview the generated YAML, and create & run the workflow.

Fail-Fast for Parallel Steps

Parallel steps now support a failFast option. When one worker fails, the remaining workers are aborted immediately instead of continuing to run.

Backstage Integration

A new Testkube plugin for Backstage is available. It provides a read-only view of environments and recent executions in your Backstage portal, with links back to the Testkube Dashboard.

Queued Execution Timeouts

Executions waiting in the queue can now be automatically aborted after a configurable timeout. Configurable at the installation level using environment variables.

Multiple GitHub App Installations

Testkube now supports multiple GitHub App installations. If your organization uses multiple GitHub organizations, all valid installations work at the same time — invalid ones are skipped.

Agent Key Rotation

The CLI now supports rotating agent secret keys. This is useful if you need to rotate the secret key for an agent.

AI Updates

Chat Overlay

A floating, resizable chat widget is now accessible from any page in the Testkube dashboard. Start a new AI conversation or switch between recent chats without leaving your current workflow. The widget persists its position and size across sessions.

AI Agent Triggers

You can now configure AI Agent triggers that automatically invoke an AI agent in response to test events. For example, set up a trigger that runs a failure-analysis agent whenever a Cypress test fails — with configurable trigger events, label selectors, trigger modes, and prompt templates.

MCP Registry Browser

Browse and install MCP servers directly from the dashboard using the new MCP Registry Browser. Search available servers, view descriptions and versions, and install them with a single click to extend your AI agents' capabilities.

Custom LLM Models

Environment administrators can now configure custom LLM providers and models for AI agents. Add your own OpenAI-compatible provider with an API key credential, register specific models, and select them in the chat model picker alongside the platform default.

Notifications for Pending Tool Approvals

A new notification center alerts you when AI agent sessions are waiting for tool approval. Review and approve pending actions directly from the notification panel without having to navigate to each chat session individually.

Default Installed Agents

New Environments now come with 4 default agents tuned for Troubleshooting, Analysis, Optimization and General Assistance.

Improvements and fixes

JUnit report attachments, including screenshots and URLs, are now rendered inline in the Report tab. Read more
Execution tags can now be edited after completion — through the Dashboard, CLI, or MCP server. Read more

Patch Release v2.7.6 (2026-03-23)

This patch release focuses on runner registration and data parsing stability.

Control Plane is available in Helm chart v2.330.6.

Fixes

Fix a panic when using PostgreSQL when listing test workflows without executions.
Remove the runner registration rate limiter to avoid unnecessary registration failures.

Patch Release v2.7.5 (2026-03-20)

This patch release focuses on PostgreSQL database migration reliability.

Control Plane is available in Helm chart v2.330.5.

Fixes

Allow out-of-order PostgreSQL migrations so release upgrades are more resilient.

Patch Release v2.7.4 (2026-03-20)

This patch release focuses on control-plane stability.

Fixes

Control Plane now emits resource lifecycle events consistently when resources are created, updated, or deleted through the Control Plane.
Control Plane now uses MongoDB v2 client.
ListExecutions MCP handler now handles empty WorkflowName and date filters correctly.
Prevent a Postgres read-write race that could affect distributed locking behavior.
Fixed Postgres empty TestWorkflows catalog.
queue-testworkflow and start-testworkflow events now correctly trigger webhooks.

Patch Release v2.7.3 (2026-03-16)

This patch release introduces fixes for Testkube CLI and Control Plane PostgreSQL BD support.

Control Plane is available in Helm chart v2.330.3.

Fixes

Include missing PostgreSQL DB table called leases.
Returning a list when requesting server to run multiple Test Workflows. As instance, using Testkube CLI to start Test Workflows by label(s) when more than one match with same label(s).

Patch Release v2.7.1 (2026-03-10)

This patch release improves webhook reliability, Agent capability visibility, and runner installation behavior.

Update both the Control Plane and the Agent to get the full set of fixes in this release. This release is available in Helm chart v2.330.1.

Fixes

queue-testworkflow and start-testworkflow events now correctly trigger webhooks.
Agent capabilities are now synced to the Control Plane based on the Agent configuration, so the UI reflects how Agents are actually configured.
runner capability is still not changed through capability sync. If you need to change it, recreate the Agent.

Monthly Release v2.7.0 (2026-03-04)

This release is available in Helm chart v2.330.0. Control Plane and Agent continue with unified versioning in 2.7.x.

What's new!

Control Plane owns Testkube Resource management

Starting in v2.7, Testkube Resource ownership moves to the Control Plane for commercial deployments. For most users, this means workflow and schedule management is now centered in Control Plane with no workflow changes required. Environment health no longer depends on continuous SuperAgent connectivity, and environments no longer automatically enter read-only mode when that connection is unavailable.

If you're upgrading from a Testkube previous version, your existing SuperAgent(s) will automatically migrate your Testkube Environment(s) with the new source-of-truth model.

Workflow sync behavior in GitOps Deployments

If your team manages Testkube Resources as CRDs, enable and validate newly introduced GitOps Agent so Git-managed CRD updates continue to reconcile as expected.

Read more about GitOps with Testkube and Testkube Agents.

warning

Breaking changes

API endpoints

The legacy execution groups endpoint was removed in v2.7.0: GET /organizations/{orgId}/environments/{environmentId}/test-workflows/{workflowId}/execution-groups

-> Use the executions list endpoint instead: GET /organizations/{orgId}/environments/{environmentId}/agent/test-workflows/{workflowId}/executions

Kubernetes lifecycle events

If you rely on Kubernetes Events or CDEvents for resource lifecycle (e.g. to trigger automation when a workflow is created from the dashboard), those will not fire for the Control Plane - originated changes mentioned above. This behavior will be fixed at the latest in the next monthly release at the end of March.

Observability updates

A large set of SuperAgent Prometheus metrics is now available in Control Plane.

AI Updates

External MCP Servers now support OAuth authentication. Read more
Many improvements and bug-fixes to the orchestration engine.

Improvements and fixes

Removed the deprecated native Slack integration configuration from Helm values: slackToken, slackSecret, and slackConfig.
Slack notifications should now be configured with Testkube Webhooks - Read more.
Extended support in JUnit visualization for non-numeric time values, such as 200ms, 2s, 0.2m, and 0.2h.
Update MongoDB image and helm chart to use latest patch version available, MongoDB version 8.2.5.
Update PostgreSQL image and helm chart to use latest patch version available, PostgreSQL version 18.2.

Patch Release v2.6.4 (2026-02-27)

This release introduces polished and improvements around our SCIM integration.

Control Plane is available in Helm chart v2.329.4.

Features and Improvements

Manage permissions in Testkube: A new toggle in Organization Management → Product features lets admins control roles, environment access, and resource group assignments directly in the Testkube dashboard — while SCIM continues to handle user provisioning and team membership automatically. Learn more about this option.

Patch Release v2.6.3 (2026-02-17)

This release focuses on stability fixes in backend and dashboard components, MongoDB index correctness for soft-deleted agents, and AI service Helm configuration improvements.

Control Plane is available in Helm chart v2.329.3 and Agent is available in Helm chart v2.6.2 (Check its release notes).

Security Updates

Updated Go modules to address reported CVEs and Docker Scout findings.
Upgraded go-jose lineage to v2.6.3 (gopkg.in/square/go-jose.v2 replacement and dependency updates) as part of the CVE remediation set.

Bug Fixes

Fixed a panic path by updating the testkube dependency to include upstream handling for nil metric timestamps.
Corrected cloud plan user quota behavior so enterprise-named plans in cloud mode no longer incorrectly appear as unlimited users.
Fixed agent uniqueness behavior with soft deletes by introducing partial unique indexes for name and secret and adding a migration to drop/recreate conflicting indexes.
Improved MongoDB migration/index handling by tolerating missing collection/index cases and tightening partial-index filter matching.
Fixed AI PostgreSQL DSN precedence in Helm: db.secretRef now takes precedence over db.url and global.postgres.dsn, and missing DB configuration now fails fast at template render time.
- Adjusted DSN resolution logic in the Testkube Enterprise Helm chart, specifically AI component, to correctly prioritize AI-specific PostgreSQL configuration over global defaults. This ensures the AI service uses its dedicated database settings when provided.
- Enhanced PostgreSQL configuration in the Testkube Enterprise Helm chart, specifically AI component, by allowing SSL mode to be configured via optional query parameters in the DSN. This provides greater flexibility for different database setups and environments.
Restored TestWorkflow YAML editor capabilities in the dashboard (schema validation, hover, autocomplete) by stabilizing Monaco YAML worker lifecycle and compatibility.
Fixed Monaco editor toolbar actions by forwarding refs through lazy-loaded Monaco wrappers.
Added guards in workflow settings to handle malformed template/workflow entities without runtime failures.

Features and Improvements

Added global Helm setting global.aiApiSubdomain in cloud API values to support AI API subdomain configuration.
Made AI service PostgreSQL SSL mode configurable when using db.secretRef by adding db.queryParams (instead of hardcoding sslmode=disable).

Monthly Release v2.6.0 (2026-02-05)

This release is available in Helm chart v2.329.0. Control Plane and Agent versions are now unified at v2.6.0.

Version Unification

Starting with this release, we've consolidated the Control Plane and Agent versioning to share the same major and minor version numbers. Patch versions may differ between components as fixes are released independently, but the major.minor version (e.g., 2.6.x) will be kept in sync, making it easier to track compatibility across your Testkube deployment.

Breaking Change

If you were using the AI Assistant in a previous release, you must now configure a PostgreSQL database for the AI service to function. The AI service uses PostgreSQL to persist conversation state and session data for the new AI Agents feature.

What's new!

AI Agents for Test Orchestration

Testkube now supports custom AI Agents that can autonomously execute complex testing tasks with human oversight. This is a major evolution from the previous AI Assistant which was limited to read-only Q&A and navigation help.

AI Agents bring true agentic capabilities to your testing platform:

Create Custom AI Agents: Define agents with custom instructions tailored to your testing workflows — from "Debug failing E2E tests" to "Generate load test reports" or "Monitor flaky test patterns."
Connect External MCP Servers: Extend agent capabilities by connecting any Model Context Protocol (MCP) server. This lets agents access external tools, documentation services, or custom integrations alongside Testkube's built-in tools.
Human-in-the-Loop Approval: Agents pause before executing sensitive operations (like creating workflows or modifying resources) and request your approval. You can approve or respond with alternative instructions.
Multi-turn Sessions: Agents maintain conversation context across multiple interactions. Continue a debugging session later, or have the agent iterate on a complex task over time.
AI Assistant Integration: The dashboard AI Assistant (now read-only) can delegate requests to AI Agent sessions.

Example Use Cases:

Automated Remediation — Analyze failed tests, identify root causes, and suggest or apply fixes automatically (see Example)
Intelligent Analysis — Detect flaky tests, analyze execution patterns, and surface anomalies (see Example)
Custom Workflows — Define agents tailored to your specific testing needs and integrate with external systems

AI Agents are available from the dashboard and can be configured per environment with granular control over which tools are enabled and which require approval before execution.

Improvements and fixes

Workflows can now be configured as silent at the Workflow level by setting spec.execution.silent to true, which will result in all executions being silent by default - Read more about Silent Workflows
On-prem customers can now configure custom LLM models via Helm instead of being locked to defaults.
AI service now uses two model types: a primary model for complex agent tasks and a lightweight model for quick tasks like generating session titles.
Default models updated to gpt-5.2 (primary) and gpt-5-mini (lightweight), with optional temperature settings for models that support it.
Control Plane scheduler improvements: synchronized MongoDB and PostgreSQL implementations, resolved race conditions during test workflow execution scheduling.
Added scheduler metrics to improve scheduler observability and monitoring - Read more about Scheduler Metrics
Fixed an issue where the member details view did not correctly show permissions inherited from resource groups.

Patch Release (2025-12-23)

Control Plane is available in Helm chart v2.327.3 and Agent is available in Helm chart v2.5.3

Control Plane v1.21.3

Fixed broken documentation references in the UI.
Fixes to SCIM server when removing admins assigned from an IDP group

Agent v2.5.3

Fix issue where executions were marked as successful if an accompanying service fails immediately.

Monthly Release v1.21.1 (2025-12-10)

This release is available in Helm chart v2.327.1.

What's new!

New Executions Dashboard! We've built a Unified Executions Dashboard that brings all your executions into one place, making debugging and monitoring significantly faster. The new dashboard provides a single view with everything you need to work efficiently:
- Custom Views & Advanced Filtering: Create private and shared views with comprehensive filter options, or use prepopulated views to quickly access common filter combinations
- Inline Details: See all relevant execution information directly in the list without clicking into detail pages
- Live Data Toggle: Control live updates to focus on current executions without interruption
With this unified view, you can spot a failing execution, view its logs with one click, and act on it immediately. No more checking workflows individually or navigating through multiple pages to get the information you need. Plus, we've added several other features to make working with executions even easier.

The Executions Dashboard is available via a dedicated icon in the left navigation toolbar.
Webhook Credentials Support — Webhooks can now securely reference Credentials, so you can inject API keys and secrets into webhook URLs, headers, or payload templates using the credentials() expression instead of hard-coding them.

This keeps your integrations with external systems (Slack, incident tools, custom endpoints, etc.) both safer and easier to maintain, since you can rotate or update secrets in one place without touching webhook definitions.
Deprecation of Tests & Suites — Over the past year, TestWorkflows has become the recommended way to orchestrate tests in Testkube — offering more flexibility, better visibility, and continuous feature improvements driven directly by user feedback.

To reduce complexity and make it easier to understand how to get started, we are now fully transitioning to TestWorkflows. Beginning with version 1.21, legacy Tests, Test Suites, Executors, and Sources are removed from the product. This ensures that all future enhancements benefit the most powerful and supported workflow engine in Testkube. Read more about the legacy deprecation here.

danger
If you are still using legacy functionality, you will need to stay on previous Testkube releases and migrate your Tests/Suites to TestWorkflows before upgrading to this release, as described in the migration guide.

Join us on Slack and we’ll guide you through the process. In the meantime, you can remain on version 1.20, which will continue receiving critical bug fixes while you complete the transition.
(Preview) Support for PostgreSQL — As part of ongoing improvements to our data layer, we are introducing PostgreSQL as a new storage option. This release includes PostgreSQL support in preview mode, which can be enabled through custom configuration documented here and here PostgreSQL brings significant benefits in performance, scalability, and data security — laying the groundwork for future enhancements.

Migration tools for existing Testkube installations using MongoDB will be provided in a future release!

Improvements and fixes

Silent executions tooltips and icon are more user friendly
Updating credential now updates the correct credential scope when there is an override
Added browser tab title to Execution Details Page

Patch Release v1.20.4 (2025-11-20)

This release is available in Helm chart v2.326.4.

Improvements and fixes

Fix SCIM compatibility issue with Microsoft Entra ID.

Patch Release v1.20.3 (2025-11-19)

This release is available in Helm chart v2.326.3.

Improvements and fixes

Fix SCIM issue where organization membership gets dropped for certain users.
Add Silent Execution icon on Execution List.
Fix broken dashboard executions link.
Replace MongoDB and MinIO Helm charts with Testkube custom charts and images.

Patch Release v1.20.2 (2025-11-13)

This release is available in Helm Chart v2.326.2.

Improvements and fixes

Insights APIs now respond faster thanks to smarter queries and better indexes.
Workflow and credential flows feel more reliable across the UI.
Deleting workflow executions is safer now that the process uses a configurable timeout (default 30m) instead of a rigid 5s timeout.
Version check now correctly checks agent version.

Monthly Release v1.20.0 (2025-11-03)

This release is available in Helm Chart 2.326.0.

What's new!

Model Context Protocol (MCP) Endpoint on Control Plane Building on the MCP Server introduced in v1.18, you can now connect AI assistants directly to dedicated MCP endpoints on your Testkube environments, no local CLI server required.

Configure your AI assistant (GitHub Copilot, Cursor, or Claude Desktop) with your environment’s MCP endpoint URL and an API key, and you’re ready to go.

Through this integration, AI agents can not only run workflows, monitor executions, analyze results, and search execution history, but also modify workflows, access resource metrics for deeper performance and resource analysis, and work directly with agents for advanced orchestration, all through natural conversation in your coding environment.

More info can be found in the MCP Server page in the docs.

Silent Executions Run workflows and tests quietly when you don’t want activity to affect your analytics or automation signals. Silent Executions are now supported directly from the UI and allow you to execute tests without persisting results in Insights, generating metrics, triggering CDEvents or webhooks, or impacting workflow health.

This is ideal for local development, debugging pipelines, experimenting with configuration changes, or running internal verification checks - without polluting dashboards, reports, or automation workflows.

More info can be found in the Silent Executions section in the Running Test Workflows docs.

Enhanced Credentials Management Managing sensitive values in Testkube just got easier. We’ve introduced a unified credentials experience with improved organization- and environment-scoped credential handling, along with dedicated management screens in both organization and environment settings.

You can now create plaintext or encrypted credentials and reference them directly in workflows using the credentials() expression, making secure secret handling both intuitive and flexible across your Testkube environments.

More info can be found in the Credentials Management page in the docs.

Monthly Release v1.19.0 (2025-10-01)

This release is available in Helm Chart 2.325.0.

What's new!

Concurrency & Queue Policies. This release adds new functionality to manage the concurrency and reliability of executions, allowing you to control the number of executions that can run simultaneously across a single or a group of workflows. This helps you ensure that the number of tests running is always in line with the capacity of both your testing infrastructure and the system-under-test.

Runner Concurrency Limits allow you to prevent overwhelming your runners with too many concurrent test executions. Whenever a concurrency threshold is met, the execution will be queued until previous executions finish. Since the execution queue might fill up due to these concurrency features or because no online and matching runners can be found, Environment Queue Limits help to prevent your workspace from getting overwhelmed. Once this limit is hit, executions will immediately abort which will also trigger its web hooks for alerting. Lastly, Execution Time-outs will abort queued and ongoing executions respectively after 1 day and 7 days with further fail-safes when runners are unable to gracefully abort.

Improvements and fixes

Improved overall filter handling to be consistent across workflows/templates and dashboard executions.
TestWorkflows: Added support for dynamic runtime environment variable overrides via the --variable CLI flag.
On-Prem: Fixed an issue where inactive SCIM users were incorrectly counted towards license quotas.
Improved Agents page. You can view and install agents with different capabilities.
Added OpenTelemetry Tracing to the Control Plane.

Architectural Improvements

The team is working on some exciting architectural updates to make Testkube more scalable and reliable. Starting from this release, the internal scheduling is changing from a push-based to a pull-based architecture. You should not notice much from this change, likely things become more reliable when the connection between the control plane and agents is flaky. Another benefit is that this sits closer to Kubernetes architecture, which will open interesting opportunities in the future. Stay tuned!

Patch Release v1.18.7 (2025-09-08)

This release is available in Helm Chart v2.324.7.

Improvements and fixes

Fixed audit log pagination bug and added missing audit event types for complete tracking coverage.
Improved handling of edge cases in the insights graph on compact screens.

Patch Release v1.18.6 (2025-09-05)

This release is available in Helm Chart v2.324.6.

Improvements and fixes

Fix issue with Status Pages not loading properly.

Patch Release v1.18.5 (2025-09-03)

This release is available in Helm Chart v2.324.5.

Improvements and fixes

Fix issue with JUnit reports not being displayed correctly.

Patch Release v1.18.4 (2025-09-01)

This release is available in Helm Chart v2.324.4.

Improvements and fixes

Minor improvements and bug fixes around SCIM server in SaaS.
Remove legacy SCIM Kubernetes resources in on-prem.

Patch Release v1.18.2 (2025-08-29)

This release is available in Helm Chart v2.324.2.

Improvements and fixes

Bug fixes for listing and editing triggers.

Monthly Release v1.18.0 (2025-08-27)

This release is available in Helm Chart 2.324.0.

What's new!

Latest release is live, bringing personalized views, streamlined enterprise integrations, and new AI assistant capabilities that let tools like GitHub Copilot and Claude directly interact with your test workflows.

Private filters let you create personalized filter sets that only you can see, separate from shared team filters. Access and manage your private filters through the new navbar section without affecting what others see.

SCIM 2.0 enterprise integration is now fully multi-tenant and available on SaaS for the first time. Configure identity provider integrations (Okta, Azure AD, etc.) directly from Organization Management > Product Features - no more Helm changes or service restarts needed for on-prem users.

New Testkube MCP Server enables AI assistants (GitHub Copilot, Cursor, Claude Desktop) to directly run and monitor Test Workflows, analyze results, search execution history, and create workflows. AI agents can now help automate debugging and identify failing patterns right from your coding environment. Learn more in the docs →

Improvements and fixes

Added dynamic filtering system that shows only available options based on current data to reduce interface clutter.
CLI: Added support for SSO login, allowing users to authenticate the Testkube CLI with their organization’s identity provider.
Agent Metrics: Expanded Prometheus metrics for Test Workflows and Execution Steps. Workflow execution metrics now include additional fields (triggered_by, tags) for richer filtering and analysis, while new metrics for step-level duration and counts provide deeper visibility into execution details.
Updated Usage & Billing page for SaaS product.

Monthly Release v1.17.1 (2025-07-29)

This release is available in Helm Chart 2.323.1.

What's new!

This release focuses on usability improvements, better observability, and more powerful reuse of workflow templates.

You can now filter workflow executions on the dashboard by name, labels, or status, making it easier to focus on relevant runs during debugging, monitoring, or audits.

Workflow Templates now support runner agent targeting directly in the template spec. This allows you to define which Runner Agents should execute your workflows once, and reuse that logic across environments and pipelines. Learn more in the docs →

Improvements and fixes

Added workflow health status to the status subresource of the Workflow object, aligning with Kubernetes best practices and improving accessibility.
Added Canceled and Not Passed events to provide better differentiation between manual stops and failed validations.
Enabled verbose logging for the Control Plane API and added debug mode for Worker Service (testkube-worker-service.enableDebugMode=true in Helm values).
Both testkube-agent and testkube-runner Helm charts are now being published to the OCI registry. They will remain accessible via the original Helm repository as well.
Simplified network configuration by using the default Ingress for WebSockets used by clients. The separate WebSocket Ingress can safely be removed without requiring other changes.

Patch Release v1.16.3 (2025-07-11)

This release is available in Helm Chart v2.322.3

Improvements

Fixed executions stuck initializing or not finishing due to race conditions caused by MongoDB Replication lag.

Doc Update (2025-07-09)

Added Workflow Cookbook on Local Development

Doc Update (2025-07-08)

Added info on disabling legacy Tests in the Agent
Added info on using private AWS Elastic Container Registry

Patch Release v1.16.2 (2025-07-02)

This release is available in Helm Chart v2.322.2

Improvements

Fixed workflows filtering by status to correctly display filtered results
Improved stability and performance of custom views for a smoother, more reliable experience
Resolved organization/environment settings tabs redirecting to incorrect tabs
Fixed browser back navigation crashes that occurred on specific application pages
Improved memory usage of Control Plane when deleting artifacts

Doc Update (2025-06-27)

Added info on custom Security Headers

Doc Update (2025-06-26)

Added example for using JMeter with Plugins

Monthly Release v1.16.1 (2025-06-26)

This release is available in Helm Chart v2.322.1.

What's new!

This release comes with a new way to deploy Runner Agents and several improvements to workflows.

You can now install a Runner Agent using only Helm. Previously a Runner Agent had to be registered using the CLI or UI prior to Helm install. This feature can be used to create Runner Agents for ephemeral environments in a single step. Documentation is available here.

Users can now create custom views to group workflows by any criteria, making it easier to organize workflows that belong together or quickly access frequently used filtered results. These views are shared between all the users of a given environment.

We've added a new Workflow Health indicator that shows a quick visual summary of a workflow's recent stability, helping users understand how reliably it's been running over time, not just surfacing flakiness, but overall consistency. The health score is calculated as PassRate x (1 - FlipRate) and shown as an icon next to each workflow in both list and detail views, using weather-style visuals to represent different health levels.

We've introduced a new Canceled status to distinguish manual cancellations from system-level Aborted executions. Previously, both were marked as Aborted. With updated icons and messaging, users can now clearly see if a test Failed, was Aborted (system error), or Canceled (manual action).

Testkube Copilot has been renamed to Testkube AI Assistant to better reflect its purpose and avoid confusion with similarly named tools. This makes it clearer that the assistant is purpose-built for the Testkube ecosystem.

Improvements and fixes

Workflows search - Revised the way search works. You can now search for workflows using multiple words separated by spaces. The search matches when all the words are contained in the workflow name.
CLI Artifact Download - Fixed authentication problems with certain cloud providers due to presence of duplicate authentication mechanism.
Executions Logs - Improved horizontal scrolling for long logs.
Insights - Fixed an issue where saving changes to an existing report saved a new one instead which led to duplicates.
Fixed an issue where schedule links would navigate to the wrong workflow.
Fixed an issue where the correct JUnit test suite icon wasn't displaying properly
Improved the timestamps displayed within execution details when workflows are still queued.
Improved handling of JUnit format when working with nested test suites
Resolved a problem with empty string default values in workflow config settings
Enhanced the agents dropdown for better usability
Fixed an issue where your selected agents weren't being preserved when running workflows
Corrected broken links in the running context that weren't properly directing to workflows
Fixed a search error that occurred when searching for empty or non-existent values in execution logs
Resolved a critical security vulnerability (F-2025-4481) related to OAuth account hijacking via redirect URLs
Strengthened security headers in our nginx configuration for better protection
Fixed an issue where the Workflow wizard shows an empty preview when combining multiple templates
Security updates in upstream dependencies.
Fix API failure when deleting expired API token.
Artifact cleaning via ENTERPRISE_ARTIFACT_RETENTION_DAYS and ENTERPRISE_ARTIFACT_MAX_STORAGE_SIZE_GB now works as expected.

Doc Update (2025-06-25)

Added docs on upcoming Workflow Health and Flakiness

Patch Release v1.15.5 (2025-06-19)

This release is available in Helm Chart v2.321.5

Improvements

Fixed an issue where the correct JUnit test suite icon wasn't displaying properly
Improved handling of JUnit format when working with nested test suites
Resolved a problem with empty string default values in workflow config settings
Enhanced the agents dropdown for better usability
Fixed an issue where your selected agents weren't being preserved when running workflows
Corrected broken links in the running context that weren't properly directing to workflows
Fixed a search error that occurred when searching for empty or non-existent values in execution logs
Resolved a critical security vulnerability (F-2025-4481) related to OAuth account hijacking via redirect URLs
Strengthened security headers in our nginx configuration for better protection

Doc Update (2025-06-19)

Added example for Integrating with Argo Events
Updated Open Source Getting Started and OpenAPI docs to the latest versions

Patch Release v1.15.4 (2025-06-06)

This release is available in Helm Chart v2.321.4

Improvements

Fixed an issue where the Workflow wizard shows an empty preview when combining multiple templates
Security updates in upstream dependencies.

Doc Update (2025-06-05)

Added example for Pushing JUnit results to Datadog

Patch Release v1.15.3 (2025-06-04)

This release is available in Helm Chart v2.321.3

Improvements

Fixed an issue where artifact previews were sometimes rendered incorrectly.
Fixed an issue where Workflow Templates could not be created due to unknown function newCredential.

Doc Update (2025-06-03)

Added example for Rerunning failed tests only with Playwright

Patch Release v1.15.2 (2025-06-02)

This release is available in Helm Chart v2.321.2.

Improvements

Fix API failure when deleting expired API token.
Artifact cleaning via ENTERPRISE_ARTIFACT_RETENTION_DAYS and ENTERPRISE_ARTIFACT_MAX_STORAGE_SIZE_GB now works as expected.

Patch Release v1.15.1 (2025-05-23)

This release is available in Helm Chart v2.321.1.

Improvements

Security updates in upstream dependencies.

Monthly Release v1.15.0 (2025-05-22)

This release is available in Helm Chart v2.321.0.

What changed?

Release Quality has been our focus for Testkube v1.15. This release is packed with a variety of bug fixes and improvements.

Our team also took a good look at the release procedures of the Testkube Control Plane:

Changelogs! Testkube's control plane now comes with public changelogs and a unified version. In the past, each service such as the dashboard, API Server, workers, etc, would have its own version. Today they all share the same version which makes it easier to communicate changes and service interoperability.
The release cadence will also be reduced to better fit enterprises. Moving forward you can expect a monthly minor release with some unplanned patch versions in between. The new process also improves predictability and stability of these releases.
Our Helm Chart is now available as an OCI Artifact. The old Helm Repository approach still works, but under the hood it will now also pull as an OCI Artifact.

Features

Added timezone to TestWorkflows cronjobs.
Added support for logging Audit Events to console.

Improvements & fixes

Fixed variety of security issues.
Added downloading of an execution's log to the dashboard.
Fixed an issue with preview URLs when changing environments.
Fixed an issue with legacy tests events for the webhook condition selector
Fixed an issue with full screen mode within execution dialogue.
Fixed a problem where workingDir were incorrectly substituted within workflow definitions.
Fixed an issue updating team and board descriptions.
Fixed workflow termination when resource metrics cannot be stored due to lack of free storage space. Improved user interface of multi-agent engine
Many other minor fixes and improvements.

Release v2.15.0 (2026-10-29)​

Upgrade notes​

AI Test Creation​

AI Governance​

AI​

Executions​

Fixes & Improvements​

Runner & CLI​

Features​

Improvements​

Fixes​

Patch Release v2.14.2 (2026-10-14)​

Upgrade notes​

Security​

Fixes​

Patch Release v2.14.1 (2026-10-05)​

Fixes​

Release v2.14.0 (2026-10-01)​

Upgrade notes​

AI Test Creation​

Workflows​

Execution Diagnostics & Insights​

Members & Audit Logs​

AI​

Operations & Improvements​

Patch Release v2.13.3 (2026-09-22)​

Improvements​

Patch Release v2.13.2 (2026-09-22)​

Fixes​

Release v2.13.1 (2026-08-27)​

Advanced GitHub Integration​

AI​

Insights & Analytics​

Operations & Security​

Improvements​

Other​

Patch Release v2.12.2 (2026-08-18)​

Improvements​

Fixes​

Patch Release v2.12.1 (2026-08-04)​

Improvements​

Fixes​

Agent​

Release v2.12.0 (2026-07-23)​

AI​

Insights & Analytics​

Operations & Security​

Improvements​

Other​

Patch Release v2.11.3 (2026-07-17)​

Fixes​

Patch Release v2.11.1 (2026-07-02)​

Security​

Monthly Release v2.11.0 (2026-06-25)​

AI Updates​

Test & Git Triggers​

Onboarding​

Insights Updates​

Operations & Security​

Runner Agent​

Patch Release v2.10.3 (2026-06-09)​

Patch Release v2.10.2 (2026-06-07)​

Improvements​

Fixes​

Monthly Release v2.10.0 (2026-06-01)​

What's new!​

Granular Test Insights​

AI Updates​

Improvements​

Fixes​

Control Plane​

Fixes​

Agent​

Fixes​

Open Source Updates​

Templated Security Contexts​

AI Analysis in the Public Execution Viewer​

Patch Release v2.9.7 (2026-05-29)​

Improvements​

Patch Release v2.9.6 (2026-05-28)​

Release v2.15.0 (2026-10-29)

Upgrade notes

AI Test Creation

AI Governance

AI

Executions

Fixes & Improvements

Runner & CLI

Features

Improvements

Fixes

Patch Release v2.14.2 (2026-10-14)

Upgrade notes

Security

Fixes

Patch Release v2.14.1 (2026-10-05)

Fixes

Release v2.14.0 (2026-10-01)

Upgrade notes

AI Test Creation

Workflows

Execution Diagnostics & Insights

Members & Audit Logs

AI

Operations & Improvements

Patch Release v2.13.3 (2026-09-22)

Improvements

Patch Release v2.13.2 (2026-09-22)

Fixes

Release v2.13.1 (2026-08-27)

Advanced GitHub Integration

AI

Insights & Analytics

Operations & Security

Improvements

Other

Patch Release v2.12.2 (2026-08-18)

Improvements

Fixes

Patch Release v2.12.1 (2026-08-04)

Improvements

Fixes

Agent

Release v2.12.0 (2026-07-23)

AI

Insights & Analytics

Operations & Security

Improvements

Other

Patch Release v2.11.3 (2026-07-17)

Fixes

Patch Release v2.11.1 (2026-07-02)

Security

Monthly Release v2.11.0 (2026-06-25)

AI Updates

Test & Git Triggers

Onboarding

Insights Updates

Operations & Security

Runner Agent

Patch Release v2.10.3 (2026-06-09)

Patch Release v2.10.2 (2026-06-07)

Improvements

Fixes

Monthly Release v2.10.0 (2026-06-01)

What's new!

Granular Test Insights

AI Updates

Improvements

Fixes

Control Plane

Fixes

Agent

Fixes

Open Source Updates

Templated Security Contexts

AI Analysis in the Public Execution Viewer

Patch Release v2.9.7 (2026-05-29)

Improvements

Patch Release v2.9.6 (2026-05-28)