What can you do with this server?

This server automates AI-powered job hunting by scanning careers pages, scoring jobs against your resume, and generating tailored application materials. * Scan Jobs (scan_jobs): Scans 130+ configured company careers pages for new job postings, scores each role against your resume using an LLM (0–100 fit score), and sends a Telegram notification with the top matches. * Draft Application (draft_application): Generates a tailored resume and cover letter for a specific job, referenced by number (e.g., #1 from the last scan) or a full job URL, saving the output files locally. * Export Jobs (export_jobs): Exports job scan results to a CSV file, with optional filtering by minimum score and date range (last N days or last scan only).

Which integrations are available for this server?

Sends notifications of top job matches and enables interactive drafting via Telegram.

mcp-autopilot-jobhunt

by tarunlnmiit

Overview Schema Related Servers Score Discussions

Python

Local

autopilot-jobhunt

Your AI job agent. Finds, scores, and drafts applications — while you sleep.

Scans 130+ company careers pages nightly → scores every role against your resume with an LLM → sends you the top matches on Telegram → drafts a tailored resume + cover letter on demand.
🔒 Drafts only — never applies. You review every draft and submit applications yourself. See PRIVACY.md for exactly what data leaves your machine.

codecov PyPI version PyPI downloads Python 3.11+ License: MIT GitHub Stars autopilot-jobhunt MCP server Listed on CodeGuilds

Seen by 90K+ people — the Instagram reel that launched it.

Published on PyPI and listed on the Official MCP Registry (io.github.tarunlnmiit/autopilot-jobhunt), Glama (Quality A), and Smithery (MCPB bundle).

📖 Full setup guide with Claude Code MCP integration → SETUP.md

⭐ Star this repo if it helps you land a job
This tool is free, open source, and runs entirely on your machine — no subscription, no credit card. The only "payment" I ask: if it surfaces a role you apply to (or land), drop a star. It takes one click, costs you nothing, and it's the single thing that pushes the project in front of the next person grinding through 130 careers pages by hand. ⭐ Star it here →

How it works

flowchart LR
    A["🌐 130+ Careers Pages"] -->|TinyFish API| B["Job Discovery"]
    B --> C["LLM Batch Scorer\n(0–100 fit score)"]
    C -->|score ≥ min| D["📱 Telegram Alert\nTop N matches"]
    C -->|on demand| E["✉️ Cover Letter\n+ Resume Bullets"]
    C --> F["📊 CSV Export"]

The scoring prompt uses your actual resume — not keywords. The LLM reads your full work history and the job description, then explains in one sentence why you fit or don't. No more guessing.

What a scan result looks like

Scanning Mistral AI...
  3 new job URLs. Fetching details...
  Scoring jobs...
  Saved 2 jobs from Mistral AI

Scanning HuggingFace...
  5 new job URLs. Fetching details...
  Scoring jobs...
  Saved 3 jobs from HuggingFace

Scanning Stripe...
  No new jobs found
...
Scan complete.
Top 5 sent to Telegram.

What the Telegram notification looks like

Job Hunt — 06 Jun 2026
5 matches found

#1 | Mistral AI | Applied AI Engineer, ML Infrastructure
📍 Paris/London/Marseille, On-site
🔧 Python, LLMs, RAG, AWS, MLOps, DevOps
✅ Role combines applied AI + ML infrastructure in EU, aligns with MLOps/RAG expertise and relocation goal
Score: 85/100  →  https://jobs.lever.co/mistral/...

#2 | HuggingFace | Staff ML Engineer
📍 Remote (EU)
🔧 Python, PyTorch, Transformers, CUDA, MLOps
✅ Open-source ML role matches deep learning and distributed training background
Score: 80/100  →  https://apply.workable.com/huggingface/...

...

Reply "apply to #N" to draft a tailored application.

Related MCP server: JobGPT MCP Server

What it does

Every night at 2:30 AM:
  ┌─────────────────────────────────────────────────────────┐
  │  Scans careers pages  →  Scores with LLM  →  Notifies  │
  │       (130+ cos)           (0–100 fit)       (Telegram) │
  └─────────────────────────────────────────────────────────┘

On demand:
  autopilot draft 1  →  tailored resume + cover letter in 60s

Usage modes

Mode 1: Standalone CLI (no Claude Code required)
  pip install autopilot-jobhunt
  autopilot scan / autopilot draft 1 / autopilot export

Mode 2: Claude Code MCP (control via natural language)
  pip install 'autopilot-jobhunt[mcp]'
  claude mcp add autopilot-jobhunt ...
  → "Scan for ML jobs" / "Draft application for job #2"

Both modes use the same config and produce the same output.

Quick start

Option A — pip install

pip install autopilot-jobhunt        # or: pip install 'autopilot-jobhunt[mcp]' for Claude Code
mkdir my-job-hunt && cd my-job-hunt
autopilot init                       # creates config.json, companies.json, resume/, .env
# Fill in config.json (API keys + your profile) and resume/YOUR_RESUME.md, then:
autopilot scan

Option B — clone (recommended if you want to customize companies or contribute)

git clone https://github.com/tarunlnmiit/autopilot-jobhunt.git
cd autopilot-jobhunt
pip install -e '.'               # standalone CLI
# pip install -e '.[mcp]'       # + Claude Code MCP integration
cp config.example.json config.json && cp .env.example .env
# Fill in your API keys and candidate profile, then:
autopilot scan

For the full walkthrough — API key setup, Claude Code MCP registration, rate limit details, and troubleshooting — see SETUP.md.

📚 Documentation

Step-by-step guides live in docs/:

Guide	Covers
Install	pip / from source / `autopilot init` scaffolding
LLM providers	OpenRouter fallback chain, Claude CLI (keyless), Anthropic API
API keys	TinyFish + OpenRouter keys, where each goes
Companies & scanning	`companies.json`, discovery + scoring, scan pacing
Integrations	Telegram notifications
MCP server & Skill	Drive the hunt from Claude Code
Config & scoring	Candidate profile, `min_score`, `top_n`
Troubleshooting	Every error we've hit, and the fix
Testing checklist	Reproducible independent verification

API keys needed

Service	Cost	Required	Where to get it
TinyFish	Free — no credit card	Always	agent.tinyfish.ai
OpenRouter	Free — 4-model fallback chain	Unless using Claude CLI / Anthropic	openrouter.ai
Telegram	Free	Optional	@BotFather on Telegram

Claude Code / MCP integration

Use autopilot-jobhunt as an MCP server inside Claude Code (CLI) or Claude Desktop.

Step 1: Install with MCP support

git clone https://github.com/tarunlnmiit/autopilot-jobhunt.git
cd autopilot-jobhunt
pip install -e '.[mcp]'

Step 2: Register with Claude Code

Option A — one command:

claude mcp add autopilot-jobhunt \
  --env TINYFISH_API_KEY=your_key \
  --env OPENROUTER_API_KEY=your_key \
  --env TELEGRAM_TOKEN=your_token \
  --env TELEGRAM_CHAT_ID=your_chat_id \
  -- python -m job_hunt.mcp_server

Option B — edit ~/.claude.json manually:

{
  "mcpServers": {
    "autopilot-jobhunt": {
      "command": "python",
      "args": ["-m", "job_hunt.mcp_server"],
      "cwd": "/absolute/path/to/autopilot-jobhunt",
      "env": {
        "TINYFISH_API_KEY": "your_key",
        "OPENROUTER_API_KEY": "your_key",
        "TELEGRAM_TOKEN": "your_token",
        "TELEGRAM_CHAT_ID": "your_chat_id"
      }
    }
  }
}

Note: cwd must point to the cloned repo — the server reads config.json and companies.json from there.

Step 3: Use it

In any Claude Code session:

"Scan for ML jobs"
"Draft an application for job #2"
"Export jobs from the last 7 days with score above 70"

Claude Desktop

Same JSON block — add it under mcpServers in Claude Desktop → Settings → Developer.

Customize your target companies

Edit companies.json. Each entry needs:

{
  "name": "Stripe",
  "careers_url": "https://stripe.com/jobs",
  "search_domain": "stripe.com",
  "location": "Remote / San Francisco, CA",
  "region": "Remote"
}

The repo ships with 130+ pre-configured EU, NZ, and remote-friendly tech companies. Add or remove as you like.

How scoring works

The LLM reads your full resume + the full job description and assigns a score 0–100:

Score	Meaning
80–100	Near-perfect fit — apply immediately
60–79	Good fit — worth applying
40–59	Partial fit — apply if pipeline is thin
< 40	Poor fit — skipped

Set min_score in config to filter. Default: 60.

Project structure

autopilot-jobhunt/
├── job_hunt/
│   ├── main.py          # CLI entry point
│   ├── scanner.py       # Job discovery + LLM scoring
│   ├── drafter.py       # Resume tailoring + cover letter
│   ├── notifier.py      # Telegram notifications
│   ├── llm_utils.py     # OpenRouter wrapper with fallback
│   ├── tools.py         # Protocol-agnostic tool layer
│   └── mcp_server.py    # MCP server (Claude/AI assistant integration)
├── demo/                # Demo scripts for recording GIF
├── resume/              # Put your resume here (gitignored)
├── state/               # Scan state (gitignored)
├── output/              # Generated applications (gitignored)
├── companies.json       # 130+ target companies
├── config.example.json  # Config template (copy to config.json — gitignored)
└── config.json          # Your config (gitignored — never committed)

LLM options

Default: OpenRouter (free)

Uses a 4-model fallback chain — all free, no credit card needed:

Model	Role
`meta-llama/llama-3.3-70b-instruct:free`	Primary — best quality
`nvidia/nemotron-3-super-120b-a12b:free`	Fallback 1 — 120B
`google/gemma-4-31b-it:free`	Fallback 2
`qwen/qwen3-coder:free`	Fallback 3

If one model hits its daily free-tier quota, the tool automatically tries the next. Zero LLM cost by default.

Alternative A: Claude Code CLI (no API key needed)

If you have Claude Code installed and authenticated, you can use it as the LLM backend — no separate API key required:

In config.json:

"llm_provider": "claude_cli"

Or via environment variable: LLM_PROVIDER=claude_cli autopilot scan

Optionally set a model: "claude_cli_model": "sonnet" (or "opus", "haiku", empty = Claude's default).

Note: Requires the claude binary in your PATH. Verify with claude --print "hi" first. The MCP server and cron jobs must run in an environment where your claude auth session is active.
Rate-limit note: Each call loads your global Claude Code context (~25–30k tokens). A nightly scan (5–15 LLM calls) burns significantly against your subscription's 7-day rate limit. Prefer OpenRouter for nightly automation; use Claude CLI for occasional on-demand drafts.

Alternative B: Anthropic API

If you have an Anthropic API key:

pip install 'autopilot-jobhunt[claude]'

In config.json:

"llm_provider": "anthropic",
"anthropic_api_key": "sk-ant-...",
"anthropic_model": "claude-haiku-4-5-20251001"

claude-haiku-4-5-20251001 is fast and cheap; claude-sonnet-4-6 gives higher quality scores. A nightly scan uses ~5–15 LLM calls total (jobs scored in batches of 10).

Contributing

See CONTRIBUTING.md. PRs welcome for:

Adding companies to companies.json
New ATS platform support (Rippling, Lever variants, Workday)
OpenAI / Gemini MCP adapters
Better scoring prompts

License

MIT — see LICENSE.

Built by @tarunlnmiit. If this saved you hours of job searching, a ⭐ means a lot.

Available Tools

3 tools

draft_applicationA

Draft a tailored resume and cover letter for a specific job.

Args:
    job_ref: Job reference — '#1' or '1' (from last scan), or a full job URL.

Returns a summary of where the output files were saved.

ParametersJSON Schema

Name	Required	Description	Default
`job_ref`	Yes

Output Schema

ParametersJSON Schema

Name	Required	Description
`result`	Yes

TDQS

A3.8/5.0

Behavior2/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

No annotations provided, so the description carries full burden. It only mentions that it drafts documents and returns a summary of file locations. It does not disclose whether files are overwritten, if the tool is destructive, or any other behavioral traits like data usage or prerequisites.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is very concise: one sentence for purpose, plus a minimal Args block. Every sentence adds value, and the important information is front-loaded.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

For a simple tool with one parameter and an output schema, the description covers the purpose, parameter usage, and return value. It is sufficient for an agent to understand the tool's function and how to invoke it.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters5/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema description coverage is 0%, but the description adds essential semantics: explains job_ref format (e.g., '#1', '1', or full URL) and its source (last scan). This fully compensates for the lack of schema documentation.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose4/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description clearly states it drafts a resume and cover letter for a specific job. It does not explicitly differentiate from sibling tools (export_jobs, scan_jobs), but the purpose is specific and unambiguous.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines3/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

The description implies usage after scanning jobs (via job_ref from last scan) but provides no explicit when-to-use or when-not-to-use guidance. It does not mention alternatives like export_jobs or scan_jobs.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

export_jobsA

Export job scan results to a CSV file in the output/ directory.

Args:
    min_score: Only include jobs with score >= this value (0 = all).
    days: Export from the last N days of history (0 = last scan only).

Returns a summary of the export.

ParametersJSON Schema

Name	Required	Description	Default
`min_score`	No
`days`	No

Output Schema

ParametersJSON Schema

Name	Required	Description
`result`	Yes

TDQS

A3.9/5.0

Behavior3/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Describes output location and return value but lacks details on file naming, overwriting behavior, or performance characteristics. With no annotations, more behavioral context could be beneficial.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

Concise description starting with the main purpose, followed by parameter details and return value. No unnecessary text.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness4/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

Given simplicity (2 params, output schema exists), description adequately explains functionality, though could mention file naming or confirmation of success.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters5/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Both parameters (min_score and days) are fully explained with default values and behavior, compensating for 0% schema description coverage.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

Clearly states the tool exports job scan results to a CSV file in the output/ directory, distinguishing it from sibling tools like draft_application and scan_jobs.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines2/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

No guidance on when to use this tool versus alternatives, nor any when-not-to-use conditions.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

scan_jobsA

Scan all configured company careers pages for new job postings, score them with AI against your resume, and send a Telegram notification with the top matches. Reads config.json and companies.json from the working directory.

ParametersJSON Schema

Name	Required	Description	Default
No parameters

Output Schema

ParametersJSON Schema

Name	Required	Description
`result`	Yes

TDQS

A4/5.0

Behavior3/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

No annotations are provided, so the description must cover behavioral aspects. It explains that it reads config files and triggers notifications, but it does not state whether the tool modifies any state, has rate limits, or is read-only. More disclosure could improve safety.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is two sentences long, with the primary action in the first sentence and supplementary detail in the second. Every word adds value, and there is no redundancy.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness4/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

Given the tool has zero parameters and an output schema exists, the description is largely complete. It covers the main workflow and dependencies, though it could hint at the output format or error handling.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters3/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

The tool has no parameters, and schema coverage is trivially 100%. The description provides context about config files but does not add parameter semantics since none exist. Baseline score of 3 is appropriate.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description clearly states the tool scans careers pages, scores jobs against a resume, and sends a Telegram notification. It distinguishes itself from sibling tools (draft_application, export_jobs) by focusing on scanning and alerting.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines4/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

The description implies the use case: to discover new job postings and receive notifications. While it doesn't explicitly contrast with siblings, the context is clear enough for an agent to infer when to use this tool.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Install Server

RayrunBuild, deploy, and operate MCP servers without becoming an infrastructure team.ray.runAd

TDQS

A4.1/5.0

Disambiguation5/5

Each tool has a clearly distinct purpose: scanning jobs, drafting applications, and exporting results. No overlap or ambiguity.

Naming Consistency5/5

All tools use a consistent verb_noun pattern with lowercase_snake_case (draft_application, export_jobs, scan_jobs).

Tool Count4/5

Three tools is on the lower end but appropriate for a focused job-hunt automation server. Each tool is necessary and well-scoped.

Completeness4/5

Covers the core workflow of scanning, drafting, and exporting. Lacks tool for direct submission or managing company config, but these are minor gaps given file-based configuration.

Maintenance

ActivityMaintained

ResponsivenessSyncing

Resources

GitHub Repository

Need Help?

Related Servers

Unclaimed servers have limited discoverability.

Looking for Admin?

If you are the server author, to access and configure the admin panel.

Related MCP Connectors

AI Applyd A
io.github.whateverneveranywhere
ATS resume scoring, job analysis, interview prep, and auto-apply that verifies each submission.
AI & Machine Learning App Automation
102
Smart Match
io.github.pkhaninejad
Analyze job listings against your resume, track applications, and generate cover letters.
RAG Systems Search AI & Machine Learning
2
keywordise A
com.keywordise
Auto-apply to jobs: matches your CV, tailors a fresh CV per posting, and applies for you.
App Automation Autonomous Agents
491
nextrole A
io.github.hyperdrift-io
A job-search companion: tailor your CV to a role, score fit, fix ATS issues. Also via MCP.
Search Workplace & Productivity
52

Related MCP Servers

Adzuna Jobs MCP Server
Search Research & Data
folathecoder
A
license
A
quality
D
maintenance
Provides AI assistants with access to the Adzuna Job Search API to search millions of job listings, analyze salary data, and research top employers across 12 countries with comprehensive filtering options.
Updated 8 months ago (2026-01-12 08:10 UTC)
7
15
MIT
JobGPT MCP Server
Browser Automation Workplace & Productivity
6figr-com
A
license
B
quality
C
maintenance
Provides tools to search & auto-apply to jobs directly on company websites, generate custom resumes, get contacts of recruiters and referrals and track applications easily
Updated 20 days ago (2026-08-15 08:22 UTC)
35
107
20
MIT
job-search-mcp
Search Web Scraping
VincenzoImp
A
license
Not graded
quality
A
maintenance
MCP server that exposes job search data from multiple boards, enabling clients to query and manage job listings via natural language.
Updated 10 days ago (2026-08-25 12:38 UTC)
7
MIT
ATS Scanner MCP Server
Browser Automation Web Scraping
jhansy29
F
license
Not graded
quality
D
maintenance
Automates ATS resume scanning via Jobscan, enabling AI to iteratively scan, analyze gaps, optimize, and rescan resumes against job descriptions to improve match rates.
Updated 6 months ago (2026-03-12 18:46 UTC)
3

Latest Blog Posts

Who's Calling? MCP Hosts Are an Identity Blind Spot (And the Spec Knows It)
By Om-Shree-0709 on July 25, 2026.
mcp
Agent Identity
OAuth 2.1
Your AI Chatbot Just Exposed Your CEO's Salary to an Intern
By Om-Shree-0709 on July 2, 2026.
Agent Identity
MCP Security
OAuth Delegation
Why MCP Servers Need Execution Sandboxing (And Why Your Current Stack Isn't Enough)
By Om-Shree-0709 on June 30, 2026.
Agentic Ai
Prompt Injection
WebAssembly

MCP directory API

We provide all the information about MCP servers via our MCP API.

curl -X GET 'https://glama.ai/api/mcp/v1/servers/tarunlnmiit/autopilot-jobhunt'

If you have feedback or need assistance with the MCP directory API, please join our Discord server