autopilot-jobhunt MCP
MCP server and CLI for job hunting: scans 130+ careers pages nightly, scores roles against your resume and drafts applications
- 218 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Your resume and job texts go to a model provider: OpenRouter, Anthropic or the Claude CLI
- Holds TinyFish and OpenRouter keys and a Telegram bot token in config.json and environment variables
- Fetches third-party sites through the TinyFish service from the company list, while never submitting applications itself
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add mcp/autopilot-jobhuntDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Your resume and job texts go to a model provider: OpenRouter, Anthropic or the Claude CLI
- Holds TinyFish and OpenRouter keys and a Telegram bot token in config.json and environment variables
- Fetches third-party sites through the TinyFish service from the company list, while never submitting applications itself
This is third-party code. Review the repository files before installing.
What it does
autopilot-jobhunt installs from PyPI and runs as a standalone CLI or as an MCP server for Claude Code. autopilot scan walks more than 130 company careers pages from companies.json and picks new postings, and a language model scores each role from 0 to 100 against the full text of your resume and explains in one sentence whether it fits. The best matches arrive in Telegram. autopilot draft 1, or a request from the agent, prepares a tailored resume and cover letter for a chosen role. Only a human sends applications: the tool creates drafts and never applies. autopilot export dumps matching jobs to CSV. The MCP server has three tools: scan_jobs, draft_application and export_jobs. The repository ships a skill, skills/autopilot-jobhunt/SKILL.md, that guides Claude Code through scanning, role selection, drafting and review and treats job descriptions as untrusted input. Scoring can use a free chain of four OpenRouter models, the Claude CLI without a key, or the Anthropic API.
Who it is for. Job seekers who follow openings across dozens of companies and want only fitting roles together with drafted applications.
Good fit when
- You wait for a role at specific companies and do not want to open their careers pages by hand every day
- You want to filter openings by the meaning of your resume rather than keywords, with the reason for each score in one sentence
- The Claude Code agent should prepare a resume and a letter on "draft an application for job number 2"
- You want the best openings in Telegram in the morning and a full CSV export weekly
- You want to edit the company list in companies.json yourself and add your own
Not a fit when
- You need automatic applications: the tool prepares drafts and sends nothing
- The resume must not go to a model provider such as OpenRouter or Anthropic
- You care about aggregators like LinkedIn and Indeed: the list is built from company careers pages
- You have no TinyFish key: it is always required to fetch pages
Example request
Scan for jobs matching my profile, show the top three, then draft a resume and a cover letter for the second oneMCP
- Transport
- stdio
- Authentication
- API key
Environment variables
TINYFISH_API_KEYrequired, secret- TinyFish key for fetching careers pages; free.
OPENROUTER_API_KEYsecret- OpenRouter key for scoring jobs. Not needed when scoring runs through the Claude CLI or the Anthropic API.
TELEGRAM_TOKENsecret- Telegram bot token for notifications.
TELEGRAM_CHAT_ID- Chat id where the bot sends the best matches.
Limitations
- A TinyFish key is always required (free, no card). Scoring needs an OpenRouter key, or the Claude CLI or the Anthropic API.
- Scan pace depends on the limits of the free TinyFish and OpenRouter keys.
- Your resume and job texts go to the model provider; the exact data list is in the repository's PRIVACY.md.
- It works from the listed companies' careers pages and does not cover aggregators. Telegram needs a bot token and a chat id.
How to disable. claude mcp remove autopilot-jobhunt takes the server out of Claude Code. pip uninstall autopilot-jobhunt removes the package. config.json, companies.json and the resume folder live in your working folder and are deleted by hand.
FAQ
Does autopilot-jobhunt submit applications by itself?
No. The tool prepares a resume and a cover letter and you submit the application. This is a fixed rule in the skill.
Which keys does it need?
A TinyFish key is always required; it is free and needs no card. Scoring needs an OpenRouter key unless you use the Claude CLI or the Anthropic API. A Telegram token and chat id are needed only for notifications.
How do I connect the server to Claude Code?
Install the package with the mcp extra and run claude mcp add autopilot-jobhunt with the needed --env flags, using autopilot mcp as the command. Then add the working folder as cwd in ~/.claude.json and check with claude mcp list.
How is the job threshold set?
In config.json: min_score drops roles below a given score and top_n caps the number of jobs in a notification.
Which models score the jobs?
A chain of four free OpenRouter models, the Claude CLI without a key or the Anthropic API all work. The options are described in the providers guide in docs.
Discussion
No questions or reviews yet.
Related
MCP serversHosted Workopia MCP server: job search in more than 90 countries, resume tailoring for a posting, cover letters and an application dashboard
Job search inside your coding agent: scores each listing against your CV from 1 to 5, tailors a PDF resume, drafts letters and tracks applications
A browser-based MCP server for job search on hh.ru: resume-matched vacancy recommendations, applications and statuses
ats-mcp-server: jobs from hh.ru and ATS platforms
ats-mcp-server
An MCP server that aggregates vacancies from hh.ru and the public Greenhouse, Ashby, Lever, Workable and Bayt APIs into one format without HTML scraping
The official Reactive Resume MCP server and skill: the agent builds and edits resumes and runs a job application tracker
A job search plugin for Claude Code: finds openings, tailors resumes, writes cover letters and fills out application forms
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| autopilot-jobhunt (this card) | no | API key | 218 | |
| JobSpy MCP | no | not required | 4,644 | searches LinkedIn, Indeed and Glassdoor but does not score by resume or write letters |
| JobPilot | no | 109 | applies to jobs by itself, whereas here a human sends the application |
README in short
The page shows the flow from page scanning to a Telegram alert and an example of scan output. It then describes two modes, a standalone CLI and MCP for Claude Code, a quick start through pip or cloning and a table of required keys. Links lead to nine guides in docs, including troubleshooting and a testing checklist.