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/sse

Wire it up

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

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

What your assistant can do

ToolWhat it does
search_jobrunr_docsHybrid BM25 + semantic search. Returns ranked pages with title, URL, section, tier (oss or pro), and a highlighted snippet.
fetch_jobrunr_docFull markdown for one page. Useful when the assistant needs to quote something verbatim.
list_jobrunr_doc_sectionsGrouped TOC for browsing before searching.
request_jobrunr_pro_trialIf 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.