Ashishkosana

jobs-mcp

Community Ashishkosana
Updated

An MCP server that gives any LLM live access to US software-engineering jobs — Greenhouse/Lever/Ashby + community feed, US-only, clearance-filtered, no API keys.

jobs-mcp

An MCP server that gives any LLM live access to US software-engineering job openings.

Point Claude (or any Model Context Protocol client)at this server and it can search current postings — pulled live from company ATSboards (Greenhouse / Lever / Ashby) and a community new-grad feed — right inside yourconversation. US-only, security-clearance / citizenship-required roles filtered out,newest first. No API keys: every source is a public endpoint.

Claude  ──calls──▶  jobs-mcp
                     • search_jobs(query, location, limit)
                     • list_sources()
                       └─▶ Greenhouse · Lever · Ashby · SimplifyJobs feed
                       └─▶ US-only · clearance/citizenship roles excluded · deduped

Tools

Tool What it does
search_jobs(query, location, limit) Live search; query words must all appear in the title (e.g. "backend engineer", "new grad software engineer"), location is "us" or a city/state substring. Returns {company, title, location, url, posted, source}.
list_sources() The ATS boards + community feed this server pulls from.

Install & connect to Claude Code

git clone https://github.com/Ashishkosana/jobs-mcp && cd jobs-mcp
python3 -m venv .venv && .venv/bin/pip install -e .

# register it (stdio transport):
claude mcp add jobs -- "$(pwd)/.venv/bin/python" -m jobs_mcp

Then just ask Claude things like "find me fresh backend software engineer roles in New York"— it calls search_jobs and answers from live data.

Works with any MCP client (Claude Desktop, etc.) — point it atpython -m jobs_mcp over stdio.

How it works

An MCP server exposes tools a client can discover and call. Here each tool is aplain Python function decorated with @server.tool(); its type hints become the inputschema the model sees. On a call, the server fetches every source concurrently (a deadboard is skipped, not fatal), applies the US + clearance filters, dedupes, and returnsstructured rows. See src/jobs_mcp/server.py.

Scope / honesty

  • Sources are a curated company list + one community feed, not "every job on theinternet" — LinkedIn/Indeed block scraping, so this uses public ATS APIs instead.
  • The clearance filter is lexical (keywords + known defense employers); it catches thecommon cases, not every phrasing.

License

MIT

MCP Server · Populars

MCP Server · New