No API key 5 live boards MIT license Python 3.9+

Give any AI agent live remote-job data — the Remote Jobs MCP server

A Model Context Protocol (MCP) server that plugs any MCP client — Claude Desktop, Cursor, Claude Code, or your own agent — directly into Remote Jobs API: live, normalized remote-job listings from 5 boards, one keyless endpoint, a skill fit-score, and a live salary band per query.

Live jobs right now

340

Source boards

Auth required

None

Tools exposed

Why an agent wants this

Most agents can already reach the web, but raw job-board pages are inconsistent, unstructured, and rate-limited. Remote Jobs API normalizes them into one stable JSON schema with parsed salaries and a skill fit-score. The MCP server wraps that endpoint as three clean tools, so an agent can ask — "what do senior Python backend roles pay remote?" — and get a structured answer in one call, no scraping, no key, no signup.

Three tools

Tool	What it returns	Example question it answers
`search_jobs`	Live jobs filtered by skills and a salary floor, ranked by fit-score	"Find senior backend Python roles paying $100k+"
`salary_band`	Min / median / max annual salary floor for a skill set	"What does a remote data engineer pay right now?"
`market_stats`	Feed health: which boards are live and how many jobs total	"How many remote jobs are on the feed this second?"

Install & connect

One dependency (the official mcp Python SDK), then point your client at the stdio server:

pip install mcp remote-jobs-api
remote-jobs-api-mcp   # launches the MCP server over stdio

Claude Desktop (claude_desktop_config.json):

{
  "mcpServers": {
    "remote-jobs": {
      "command": "remote-jobs-api-mcp",
      "env": { "RJA_BASE_URL": "https://remote-jobs-api.tten.no" }
    }
  }
}

Any MCP client that spawns a stdio server (Cursor, Claude Code, OpenRouter harnesses, custom agents) uses the same entry point. Override RJA_BASE_URL to self-host; the default is the free public, keyless instance.

What an agent actually gets (live this page)

From search_jobs("python,backend", min_salary=100000) — real listings, live at page build time:

[
  {"title": "Senior Software Engineer Case Execution", "company": "Pivotal Health", "fit_score": 30, "salary_min": 190000},
  {"title": "Staff Software Development Engineer SDM", "company": "Delinea", "fit_score": 15, "salary_min": 150000},
  {"title": "Senior Director, Community", "company": "HackerOne", "fit_score": 0, "salary_min": 247000}
]

fit_score is how well a listing matches the skills you passed (0–100); the feed also discloses structured salary_min/salary_max so salary questions are answerable, not guesswork.

Under the hood

One endpoint: GET https://remote-jobs-api.tten.no/v1/jobs?skills=...&min_salary=... — no auth, no quota surprises.
Same schema as the REST API and the remote-jobs-export CLI — so agents, scripts, and humans see identical data.
Stdlib-only data layer (urllib); the only third-party import is the MCP SDK. Self-host the API in one docker compose up if you want it on your own infra.

Self-host (optional) — the MCP server calls a base URL, so you can run your own feed and point any client at it:

git clone https://github.com/earnnova-dev/remote-jobs-api
cd remote-jobs-api
docker compose up -d      # serves the API
RJA_BASE_URL=http://localhost:8000 remote-jobs-api-mcp