Skip to main content
Testkube 2.13.0 is out! New Advanced GitHub Integration, test case level flakiness, and much more! Read More

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 v2.13.1 (2026-08-27)​

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

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)​

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​

  • Login and signup are now separate pages.

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 TestWorflow 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.

Read more: Viewing Open Source Executions

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.

Read more: Viewing Open Source Executions

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 keepalived signal to avoid timeouts while following long Test Workflow executions from CLI and integration plaforms.
  • 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!​

Login via Google​

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.

Read more: Credentials — Vault Backend Setup

Resource Group Credentials​

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

Read more: Credentials

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.

Read more: Workflow Marketplace

Share Data Between Steps​

Steps can now pass values and files to each other — write outputs or drop files in one step, read them in any later step.

Read more: Sharing Data Between Steps

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.

Read more: Parallel Steps

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.

Read more: Backstage Integration

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.

Read more: Testworkflow Timeouts

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.

Read more: GitHub Apps with Testkube

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.

Read more: Agent Key Rotation

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.

Read more: AI Agent Chats

Chat Sidebar​

A persistent side panel lets you work with AI assistance alongside your regular dashboard workflow. The sidebar supports multiple simultaneous chat tabs, each with independent conversation state. Panel width is resizable and remembered across sessions.

Read more: AI Agent Chats

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.

Read more: AI Agent Triggers

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.

Read more: Connected MCP Servers

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.

Read more: Configuring AI Models

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.

Read more: AI Agent Chats - MCP Tool Approval

Default Installed Agents​

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

Read more: Defining AI Agents

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.

Read more: Testkube Resource Management

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.

Read more: Control Plane Metrics

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.

Read more about the AI Functionality in Testkube

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.

Read more about the AI Agents feature

New MCP Server Tools

The Testkube MCP Server now includes powerful new tools for querying and exploring your test data:

  • query_workflows and query_executions: Use JSONPath expressions to extract specific data from multiple workflows or executions at once. Query patterns like $..image to find all container images across workflows, or $.result.steps.* to analyze step results across executions.
  • get_workflow_schema and get_execution_schema: Retrieve the full YAML schema for TestWorkflow definitions and execution data. These tools help AI assistants understand the complete data structure when creating, updating, or analyzing workflows and results.

These tools make it easier for AI assistants to perform bulk analysis and generate insights across your test infrastructure.

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.

Read more about Concurrency & Queueing

Listener Agents. This release introduces the concept of Listener Agents that allow you to listen for TestTriggers in multiple namespaces/clusters within a single Testkube Environment. We added, a new selector field for Test Triggers enables agent-based filtering so you can filter out events from specific Agents and run triggers only in selected environments.

Under the hood, this release introduces a new capabilities model for the Testkube Agent that gives you granular control over how agents operate. Newly created agents can register with runner and/or listener capabilities to fit your architecture.

Read more about Agents

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)​

Doc Update (2025-07-08)​

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)​

Doc Update (2025-06-26)​

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)​

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)​

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 whe combining multiple templates
  • Security updates in upstream dependencies.

Doc Update (2025-06-05)​

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)​

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.