Case Study · Open Source · 100+ GitHub Stars

Autopilot Job Hunt

An autonomous AI agent that hunts jobs while you sleep — pip install autopilot-jobhunt. It scans 130+ company careers pages nightly, scores every role 0–100 against your full resume, and pushes the best matches to Telegram before breakfast.

100+ GitHub stars29 forksMIT LicensePyPIMCP server on Glama
⎋ View on GitHub📦 PyPIDiscuss this project

01Why it exists

Job hunting is a nightly grind of opening the same careers pages, skimming the same irrelevant listings, and guessing whether a role actually fits. Keyword alerts are noise: they match words, not careers. The interesting question is "how well does this whole role match my whole resume?" — and that is an LLM question, not a keyword question.

02See it run

Recorded against the bundled demo fixtures — the exact autopilot scan → draft → export flow:

03How it works

  1. Nightly scanAn autonomous agent walks 130+ company careers pages on a schedule — not job boards, the primary sources — and normalises every posting.
  2. LLM fit scoringEvery role is scored 0–100 against the full resume, not keywords. A 4-model OpenRouter fallback chain keeps the pipeline running at zero API cost.
  3. Telegram deliveryTop matches land in Telegram with score and reasoning. No dashboard to remember to check.
  4. On-demand documentsOne command drafts a role-tailored resume + cover letter in about 60 seconds.
  5. MCP modeRuns as an MCP server (publicly listed on Glama) — the whole pipeline drives from natural language inside Claude Code or Claude Desktop.

04Results

100+GitHub stars
130+careers pages nightly
0–100LLM fit score
$0LLM API cost

05Stack

PythonOpenRouter (4-model fallback)Claude MCPTelegram Bot APITinyFish APIDockerGitHub ActionsPyPIRaspberry Pi