Use JobRunr from inside your AI assistant

Your AI assistant has opinions about JobRunr. Some of them are wrong: invented method signatures, configuration properties spelled three different ways, Pro features described as if they were OSS. You’ve probably worked around it more than once.

The JobRunr MCP server fixes that. It serves the full jobrunr.io documentation as a hybrid search index (BM25 plus ONNX MiniLM embeddings) over the Model Context Protocol. Point your editor at one URL and the assistant searches the actual docs instead of guessing.

Endpoint: https://mcp.jobrunr.io/mcp

Wire it up

claude mcp add --transport http jobrunr-docs https://mcp.jobrunr.io/mcp

Restart Claude Code, then /mcp to confirm. Ask it anything about JobRunr and search_jobrunr_docs fires on its own.

Add to ~/.cursor/mcp.json:

{
  "mcpServers": {
    "jobrunr-docs": {
      "url": "https://mcp.jobrunr.io/mcp"
    }
  }
}

Restart Cursor. The four JobRunr tools appear under Settings → MCP.

Same URL, whichever MCP config UI your client exposes:

https://mcp.jobrunr.io/mcp

A one-line url entry is usually enough. Restart after saving.

For any other MCP client, or to inspect the JSON-RPC traffic:

npx @modelcontextprotocol/inspector --cli https://mcp.jobrunr.io/mcp --transport http --method tools/list

Drop the --cli flags for the browser UI: transport Streamable HTTP, URL https://mcp.jobrunr.io/mcp, click Connect. You’ll see the four tools, their schemas, and can call them by hand.

Note
Added the server before August 2026? You’ll have it on the older https://mcp.jobrunr.io/sse URL. That endpoint still works, but it uses a transport the Model Context Protocol has since deprecated. Re-add the server with the URL above to move to Streamable HTTP — in Claude Code that’s claude mcp remove jobrunr-docs followed by the command in the tab above.

What your assistant can do

Tool	What it does
`search_jobrunr_docs`	Hybrid BM25 + semantic search. Returns ranked pages with title, URL, section, tier (`oss` or `pro`), and a highlighted snippet.
`fetch_jobrunr_doc`	Full markdown for one page. Useful when the assistant needs to quote something verbatim.
`list_jobrunr_doc_sections`	Grouped TOC for browsing before searching.
`request_jobrunr_pro_trial`	If a Pro feature comes up, your agent can request a free trial for you. It asks for your work email and company first. Nothing fires until you say go.

The server doesn’t generate answers. It surfaces relevant pages and lets your assistant synthesize.

Updates and health

The index rebuilds on every jobrunr.io deploy and stays in sync within minutes. Nothing for you to update.

If you ever want a liveness check:

curl https://mcp.jobrunr.io/actuator/health
# {"status":"UP",...}

No MCP client?

Drop /llms.txt into any LLM context. Same docs, curated as a link index.

Source

The server is open source: github.com/jobrunr/docs-mcp-server. Issues and PRs welcome.