jmeter-docs-mcp
v0.1.0
Published
MCP server and CLI that gives AI agents access to the official Apache JMeter user manual, component and function reference, and .jmx test plan inspection.
Maintainers
Readme
jmeter-docs-mcp
MCP server and CLI that gives AI agents access to the official Apache JMeter user manual and inspects .jmx test plans.
Community project, not affiliated with the Apache Software Foundation.
Run as an MCP server
With no command, the binary starts an MCP server over stdio:
npx -y jmeter-docs-mcp@latestTools: jmeter_docs_search, jmeter_reference_lookup, jmeter_docs_fetch, jmeter_inspect_test_plan. See the repository README for client setup.
CLI
The package installs two equivalent binaries, jmeter-docs-mcp and jmeter-docs.
| Command | Description |
|---------|-------------|
| serve | Start the MCP server over stdio (default) |
| search <query...> [-n, --limit <n>] [--json] | Search the user manual |
| lookup <name...> [-k, --kind component\|function\|property] [--max-chars <n>] [--json] | Show the reference entry for a component, function, or property |
| fetch <url> [-s, --section <heading>] [--max-chars <n>] [--json] | Fetch a jmeter.apache.org page, or one section, as markdown. Accepts page names such as functions.html#__time. |
| sections [page] | List indexed sections, optionally for one page |
| inspect <file.jmx> [--json] | Outline a test plan and report best-practice findings |
| clear-cache | Delete cached documentation pages |
Examples:
jmeter-docs search "distributed testing firewall ports"
jmeter-docs lookup "JSON Extractor"
jmeter-docs lookup jmeter.save.saveservice.output_format --kind property
jmeter-docs fetch component_reference.html#HTTP_Request_Defaults
jmeter-docs inspect ./load-test.jmx --jsonHow it works
- Fetches the 27 user manual pages from jmeter.apache.org, 6 at a time, the first time a search or lookup runs. Pages are cached on disk (see
JMETER_DOCS_CACHE_DIR) and refreshed afterJMETER_DOCS_CACHE_TTL_HOURS(default 168). A stale copy is used if the site is unreachable. - Converts each page to markdown and splits it at
h1/h2, so every component and function is its own section with a deep link. - Ranks sections with BM25, weighting titles and breadcrumbs above body text.
jmeter_docs_fetchonly fetches fromjmeter.apache.org.inspectparses the JMX locally and never executes anything.
Development
npm install
npm run build
npm testRequires Node.js 22+.
