M
MCP Directory CURATED DIRECTORY
911 MCP servers indexed
中文
← Back to directory
T

Testkube MCP Server

Official
MCP server for the open testing platform built for AI-driven engineering teams
How to install ↓ Compare GitHub source repository ↗
Category
Dev Tools #74 of 438
Stars
★ 1.7k Very popular
Transport
stdio (local process) · Streamable HTTP
Runtime
Docker · Hosted (no local runtime)
Credentials
API key / credential required
License
Other / unspecified
Last commit
63FMRS · C

Testkube MCP Server is Testkube's official MCP entry point for AI-driven engineering teams running automated tests on Kubernetes, covering test workflows, executions, and artifacts. It supports both a locally run OCI image over stdio and a hosted streamable-http remote endpoint, gated by a Testkube access token plus organization and environment credentials, with documentation at https://docs.testkube.io/articles/mcp-overview. The source does not enumerate individual tools, so the tool list is empty.

Strongest · Maintenance 15/20 Weakest · Reliability 9/20

Reliability
9/20
Security and permissions
13/20
Maintenance
15/20
Documentation
13/20
Setup experience
13/20
Why each score
Reliability 9/20
The manifest declares two transports (stdio via the OCI image docker.io/kubeshop/mcp-server:latest and a streamable-http remote) with controllable knobs (TK_MCP_TRANSPORT, SHTTP host/port, TK_DEBUG), and the required dependencies (org/environment IDs, access token, control-plane URL) are explicit, so the start path and diagnosability are at least defined. However, the supplied material contains only server.json, the README, and repository metadata — no committed tests and no CI workflow exercising the MCP init + tool-list handshake — so declared tools cannot be statically verified against real behavior, and no tool inventory, timeout, retry, or failure semantics are given. Static calibration caps reliability at 12; given the missing execution evidence, it drops to 9.
Security and permissions 13/20
Credential handling looks sound: TK_ACCESS_TOKEN is marked isSecret: true, the remote URL requires organization/environment variables, the default control-plane endpoint is HTTPS (api.testkube.io), and no real tokens or secrets appear in the README or examples. The scope (running tests, reading execution results and artifacts) matches the product's stated purpose, and no malware, covert exfiltration, or irreversible delete/pay/deploy defaults are visible. Deductions: the manifest does not state whether mutating operations (triggering or deleting executions, editing workflows) require human confirmation, nor does it recommend least-privilege token scopes; the remote endpoint's authentication scheme is undescribed; and TK_CONTROL_PLANE_URL / TK_DASHBOARD_URL are overridable to arbitrary hosts without data-flow disclosure. No red line is triggered, so this is not scored 0-4.
Maintenance 15/20
The project is active and not archived: 1,656 stars with only 49 open issues, manifest version 1.0.1 published against the official 2025-12-11 schema, plus in-repo governance files (CONTRIBUTING.md, ARCHITECTURE.md, DEVELOPMENT.md). Deductions: the repository license field is NOASSERTION while the manifest claims MIT, so the license must be human-verified (not absent, hence not 0); no evidence of dependency-update automation or a security-response channel (e.g. SECURITY.md); and no reviewable release cadence or changelog in the supplied material.
Documentation 13/20
Documentation resources are substantial: the manifest lists a homepage and dedicated MCP documentation page, and the README is layered, covering the two deployment models (OSS agent vs. commercial control plane), triggering options, and result/artifact aggregation, with links to focused docs. Deductions: MCP-specific installation, authentication, tool parameters, permission scopes, rate limits, cost, and troubleshooting are only linked, not described in the visible text, so they cannot be statically verified; no tool inventory or parameter schema is listed; enterprise/trial links carry marketing parameters that readers must weigh; and no minimal MCP usage example is included. Scored in the 10-13 band at 13.
Setup experience 13/20
The install paths are explicit and multiple: stdio via the OCI image docker.io/kubeshop/mcp-server:latest with environment variables (TK_ACCESS_TOKEN, TK_ORG_ID, TK_ENV_ID required; others optional with defaults), plus a hosted streamable-http remote, and the README points to Helm/CLI install and a Quickstart. Deductions: no client configuration examples for mainstream MCP clients (Claude Desktop, VS Code, etc.) appear in the supplied files; a Testkube org/environment and issued token must exist first (external prerequisite); the OCI package uses the mutable latest tag rather than a pinned digest, hurting reproducibility; and self-hosted shttp requires choosing host/port. Static calibration caps setup at 15; without committed verifiable execution evidence it drops to 13.

Static review · not runListed 2026-09-10

Read the FMRS scoring method →

Fit and risk

What it can accessUses the network

Best for

  • Engineering teams already running Testkube on Kubernetes for automated testing
  • Platform and DevOps teams that want to manage test workflows and executions with AI assistants
  • Organizations looking to fold test result inspection and troubleshooting into AI workflows

Not for

  • Users who have not deployed Testkube or lack organization/environment credentials
  • Teams unwilling to expose a test access token to an AI client
  • General-purpose testing scenarios unrelated to Testkube or non-Kubernetes testing

Required permissions

  • Read and use the Testkube access token (TK_ACCESS_TOKEN) for authentication
  • Access test workflows, executions, and artifacts within the specified Testkube organization and environment
  • Remote mode requires network access to https://api.testkube.io and the organization/environment path

Risks and side effects

  • The access token is a sensitive credential; leaking it could let others read or act on test data in the target organization/environment
  • AI assistants may trigger or affect real test executions and the Kubernetes test resources they consume
  • The remote streamable-http mode sends requests and credentials to the Testkube cloud control plane
  • Overly broad token permissions or wrong organization/environment IDs could reach unintended environments

Setup

Before you start

Runtime：Docker · Hosted (no local runtime)

TK_ACCESS_TOKEN requiredsecret Testkube access token for authentication, obtained from the Testkube console; secret.
TK_ORG_ID required Testkube organization ID (e.g., tkcorg_xxxxxxx), found in the Testkube console.
TK_ENV_ID required Testkube environment ID (e.g., tkcenv_xxxxxxx), found in the Testkube console.
Other optional settings (6)
TK_CONTROL_PLANE_URL optional Control Plane URL, defaults to https://api.testkube.io; change for self-hosted instances.
TK_DASHBOARD_URL optional Testkube Dashboard URL, optional.
TK_DEBUG optional Enables debug mode, default false.
TK_MCP_TRANSPORT optional MCP transport type: stdio or shttp, default stdio.
TK_MCP_SHTTP_HOST optional Host to bind when using the shttp transport, optional.
TK_MCP_SHTTP_PORT optional Port to bind when using the shttp transport, optional.
  1. Obtain a Testkube access token, organization ID (tkcorg_xxxxxxx), and environment ID (tkcenv_xxxxxxx) from Testkube.
  2. For local stdio: configure your MCP client to run the OCI image docker.io/kubeshop/mcp-server:latest and set the required environment variables TK_ACCESS_TOKEN, TK_ORG_ID, and TK_ENV_ID; optionally set TK_CONTROL_PLANE_URL (default https://api.testkube.io), TK_DASHBOARD_URL, TK_DEBUG, TK_MCP_TRANSPORT (stdio or shttp, default stdio), TK_MCP_SHTTP_HOST, and TK_MCP_SHTTP_PORT.
  3. For the remote option: connect your client to the streamable-http endpoint https://api.testkube.io/organizations/{organization_id}/environments/{environment_id}/mcp, substituting your real organization and environment IDs.
  4. Consult the official documentation at https://docs.testkube.io/articles/mcp-overview for client-specific configuration details.

Check that it works

Connect your client to https://api.testkube.io/organizations/{organization_id}/environments/{environment_id}/mcp, confirm Testkube tools appear in the client's tool list, or ask it to list workflows to prove the connection works.

Troubleshooting

  1. Verify that TK_ACCESS_TOKEN, TK_ORG_ID, and TK_ENV_ID are set correctly and are not empty
  2. Confirm organization_id and environment_id match the values in Testkube (tkcorg_ and tkcenv_ prefixes)
  3. For remote mode, check network access to https://api.testkube.io; for local mode, check the control plane URL (default https://api.testkube.io)
  4. Enable TK_DEBUG for more detailed debug output
  5. When using the shttp transport, verify TK_MCP_TRANSPORT, TK_MCP_SHTTP_HOST, and TK_MCP_SHTTP_PORT

Things to try

Once connected, you can ask your AI assistant things like:

  • List all test workflows in my Testkube environment
  • Run the test workflow that failed most recently and show me the results
  • Show me the artifacts from the last test execution

Use cases

Trigger and monitor Testkube test workflow executions from an AI assistant
Query test execution results, logs, and artifacts for centralized troubleshooting
Integrate test orchestration into AI-driven engineering workflows on Kubernetes
Provide a natural-language entry point to test results in CI/CD and GitOps setups

Supported clients

Claude DesktopPartial support

Listed from the project's documentation, not tested by this site.

Overview

Testkube MCP Server is maintained by Testkube and lets AI assistants manage Testkube test workflows, executions, and artifacts. It is available in two forms: a locally run OCI image, docker.io/kubeshop/mcp-server:latest, using stdio transport, and a hosted streamable-http remote at https://api.testkube.io/organizations/{organization_id}/environments/{environment_id}/mcp. The remote endpoint requires organization_id (for example tkcorg_xxxxxxx) and environment_id (for example tkcenv_xxxxxxx). The local image requires TK_ACCESS_TOKEN, TK_ORG_ID, and TK_ENV_ID, with optional TK_CONTROL_PLANE_URL, TK_DASHBOARD_URL, TK_DEBUG, TK_MCP_TRANSPORT, TK_MCP_SHTTP_HOST, and TK_MCP_SHTTP_PORT. It suits teams running automated tests on Kubernetes that want to trigger tests, inspect results and artifacts, and troubleshoot from an AI assistant. It is not suitable for users without a Testkube organization/environment or those unwilling to give an AI client access to a test access token.

Similar servers

Context7 80 · B

Upstash's official server providing up-to-date third-party library docs for AI coding assistants

★ 63.3k · Tools 2 Compare with this →
Compare with other MCP servers

Source revision 2bf296380ba3 Data synced 2026-10-19 Read the FMRS scoring method