test-plan
This server does not currently have a deployable release on Glama. Follow the repository's setup instructions to install and configure it for your MCP client, including any required credentials and local dependencies.
Test plan
A small local server that test harnesses report results into, and that agents can read and drive over MCP.
One Node.js process serves:
HTML pages for people, with no JavaScript: a summary of pass, fail and not-run counts by group, and a page per test with its plan and history.
A REST API for test harnesses.
An MCP endpoint at
/mcpfor agents.
Data lives in SQLite. There are no npm dependencies; it uses Node 24's built-in node:http and node:sqlite. It is for local use and has no auth.
This repository is also a Claude Code plugin that connects Claude to the server and teaches it how to use it.
Run the server
docker run -d --name test-plan --restart unless-stopped -p 127.0.0.1:3000:3000 -v test-plan-data:/data ghcr.io/kharabasz/test-plan:latestOpen http://localhost:3000. Data stays in the test-plan-data volume.
From a clone, docker compose up -d --build does the same with a locally built image, and npm start runs it without Docker (Node 24 or later, database at ./data/testplan.db).
Settings are environment variables: PORT (default 3000), HOST (default 127.0.0.1, 0.0.0.0 in the container) and DB_PATH.
Related MCP server: Testing Agent MCP Server
Connect an agent
Claude Code plugin
/plugin install test-plan --marketplace kharabasz/test-planOr from a shell:
claude plugin marketplace add kharabasz/test-planclaude plugin install test-plan@test-planThe plugin adds the test-plan MCP server and a test-plan skill that covers starting the server, the tools, and writing harness reporters. It connects to http://localhost:3000/mcp; set TEST_PLAN_URL before starting Claude Code to point it elsewhere.
Any other MCP client
Point it at the Streamable HTTP endpoint http://localhost:3000/mcp. For Claude Code without the plugin:
claude mcp add --transport http test-plan http://localhost:3000/mcpMCP tools
Tool | What it does |
| Counts for the plan, each group, sub group and test, using each test's latest result |
| Tests whose latest result failed, with the failure text and page link |
| Read and add groups |
| Read, add and edit tests |
| Soft delete a test and undo it |
| A test's history, newest first |
| Append a pass or fail; |
The endpoint is stateless, answers with JSON (no SSE stream), and refuses requests whose Origin is not local.
Model
Test group:
id,description, optionalparentId. A group with a parent is a sub group, and a sub group cannot have sub groups of its own.Test:
id,groupId,shortDescription,fullDescription(the plan). Tests can be soft deleted and restored. Deleted tests are left out of the counts but their pages and history stay.Test result:
testId,runAt,status(passorfail), optionalfailureDescription. Results are append only; nothing can change them and the database refuses updates and deletes.
runAt is an ISO 8601 date-time with an offset, such as 2026-11-16T09:30:00+01:00 or 2026-11-16T08:30:00Z. It is stored and shown as sent. The latest result of a test is the one with the latest instant; on a tie, the one reported last wins.
REST API
All bodies are JSON. Errors come back as { "error": "..." } with a 4xx status.
Method | Path | Body | Result |
GET |
|
| |
GET |
| counts by group, sub group and test | |
GET |
| tests whose latest result failed | |
GET |
| all groups | |
POST |
|
| 201 group; 409 if the id exists; 400 if the parent is missing or is itself a sub group |
GET |
| group | |
GET |
| tests, deleted ones only with | |
POST |
|
| 201 test; 409 if the id exists |
GET |
| test with | |
PATCH |
|
| updated test |
DELETE |
| soft deletes, returns the test | |
POST |
| undeletes, returns the test | |
GET |
| results, newest first | |
POST |
|
| 201 result |
Ids are free text up to 200 characters; URL-encode them in paths. A failureDescription is only accepted with status: "fail". Results can be reported for a deleted test; they are kept but not counted while the test is deleted.
A harness that runs every time can create its groups and tests on each run and treat 409 as "already there".
curl -s -X POST localhost:3000/api/groups -H 'content-type: application/json' -d '{"id":"checkout","description":"Checkout"}'curl -s -X POST localhost:3000/api/tests -H 'content-type: application/json' -d '{"id":"CO-001","groupId":"checkout","shortDescription":"Guest checkout","fullDescription":"1. Add an item\n2. Check out as a guest\nExpected: confirmation page"}'curl -s -X POST localhost:3000/api/results -H 'content-type: application/json' -d '{"testId":"CO-001","runAt":"2026-11-16T09:30:00+01:00","status":"fail","failureDescription":"Got HTTP 502 from /api/orders"}'Pages
/lists the top-level groups with pass, fail and not-run counts. Expand a group to see its sub groups and tests, and a sub group to see its tests./tests/{id}shows a test's plan and history, newest first. Expand a failed run to read the failure description.
Development
npm testCI runs the tests and claude plugin validate, then publishes a multi-arch image to ghcr.io/kharabasz/test-plan from main (latest) and from v* tags.
License
MIT
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
This server cannot be deployed
Maintenance
Related MCP Connectors
Free test management over MCP: projects, test cases, plans, runs, environments and tasks.
Read-only lookup of tested agent tools and MCP servers: scores, test evidence, deprecation.
Read and write Mission Control state via MCP — projects, tasks, subtasks, templates, status updates.
Start and manage Reason cloud coding agents, projects and sessions from any MCP client.
Related MCP Servers
- AlicenseAqualityDmaintenanceEnables test execution and management through MCP clients, allowing retrieval of projects, listing tests, executing tests with browser selection, and monitoring results.634 npmMIT
- FlicenseNot gradedqualityDmaintenanceProvides self-healing UI/API test automation via Plan-Generate-Run-Heal cycle, exposing MCP tools for test platforms to trigger batches and retrieve reports.-
- AlicenseNot gradedqualityBmaintenanceEnables pulling test data, inspecting telemetry channels, running analysis, and writing results back from any MCP client.MIT
- AlicenseNot gradedqualityCmaintenanceEnables running YAML-defined test suites against MCP servers and agent workflows, with live or simulated execution, assertions, and regression reporting.72 npmMIT