handshake-mcp-server
Enables AI assistants to connect to Handshake to search jobs, browse employers, explore events, and pull student or employer profiles.
README
Handshake MCP Server
<p align="left"> <a href="https://pypi.org/project/handshake-mcp-server/" target="_blank"><img src="https://img.shields.io/pypi/v/handshake-mcp-server?color=blue" alt="PyPI"></a> <a href="https://github.com/sudhxnva/handshake-mcp-server/actions/workflows/ci.yml" target="_blank"><img src="https://github.com/sudhxnva/handshake-mcp-server/actions/workflows/ci.yml/badge.svg?branch=master" alt="CI Status"></a> <a href="https://github.com/sudhxnva/handshake-mcp-server/actions/workflows/publish.yml" target="_blank"><img src="https://github.com/sudhxnva/handshake-mcp-server/actions/workflows/publish.yml/badge.svg" alt="Publish"></a> <a href="https://github.com/sudhxnva/handshake-mcp-server/blob/master/LICENSE" target="_blank"><img src="https://img.shields.io/badge/License-Apache%202.0-brightgreen?labelColor=32383f" alt="License"></a> </p>
Through this Handshake MCP server, AI assistants like Claude can connect to your Handshake account — the leading job and internship platform for students. Search jobs, browse employers, explore events, and pull student or employer profiles, all from your AI chat.
Installation Methods
<!-- TODO: Add demo video here -->
Usage Examples
Find software engineering internships at companies in San Francisco
What events are coming up on Handshake this week?
Get details for job posting 12345678 — does it sponsor visas?
Show me the employer profile for Google on Handshake
Features & Tools
| Tool | Description | Status |
|---|---|---|
get_student_profile |
Get a student's profile by user ID (education, experience, skills, etc.) | Working |
get_employer_profile |
Get employer overview, open jobs, and reviews | Working |
search_employers |
Search for employers by keyword | Working |
get_job_details |
Get full job/internship posting with metadata (salary, visa, dates) | Working |
search_jobs |
Search jobs with keyword, location, type, and sort filters | Working |
get_event_details |
Get event details for career fairs and info sessions | Working |
search_events |
Search for upcoming events | Working |
close_session |
Close the browser session and free resources | Working |
[!IMPORTANT] Handshake is a student-gated platform — you must have a valid student account and log in before tools will work. Run
uvx handshake-mcp-server --login(or the setup wizard) to authenticate.
<br/>
🚀 uvx Setup (Recommended)
Prerequisites: Install uv and a Handshake student account.
1. Install the Patchright Chromium browser (one-time)
uvx --from handshake-mcp-server patchright install chromium
2. Run the setup wizard
uvx handshake-mcp-server setup
The wizard asks whether you want Docker or local mode, handles login, and prints the exact claude mcp add-json command to register the server with your MCP client.
3. Register with your MCP client
After setup, paste the command the wizard printed. For example:
# Local mode (opens a browser window on your machine)
claude mcp add-json handshake '{"command":"uvx","args":["handshake-mcp-server"]}'
# Docker mode (runs headless in a container)
claude mcp add-json handshake '{"command":"uvx","args":["handshake-mcp-server","docker"]}'
Or add manually to your MCP client config (claude_desktop_config.json or equivalent):
{
"mcpServers": {
"handshake": {
"command": "uvx",
"args": ["handshake-mcp-server"]
}
}
}
Restart your MCP client. Done.
[!NOTE] The server keeps a single Chromium browser open for its entire lifetime — this avoids re-authentication overhead and makes subsequent tool calls much faster. With
--no-headless, a browser window stays visible for the duration of the session. This is normal.
uvx Setup Help
<details> <summary><b>🔧 CLI Options</b></summary>
| Flag | Description |
|---|---|
--login |
Open browser for manual login, save profile, then exit |
--logout |
Clear saved authentication profile and exit |
--status |
Check current session status and exit |
--no-headless |
Run browser in headed (visible) mode |
--virtual-display |
Run via Xvfb virtual framebuffer (Linux only) |
--vnc-login |
Start a noVNC web server for browser-based login |
--vnc-port |
Port for noVNC server (default: 6080) |
--transport |
MCP transport: stdio (default) or streamable-http |
--host |
HTTP host (default: 127.0.0.1) |
--port |
HTTP port (default: 8000) |
--log-level |
DEBUG, INFO, WARNING, or ERROR (default: WARNING) |
HTTP mode example:
uvx handshake-mcp-server --transport streamable-http --host 127.0.0.1 --port 8000
</details>
<details> <summary><b>❗ Troubleshooting</b></summary>
Login issues:
- Make sure you have a valid Handshake student account
- Handshake may show a CAPTCHA on first login —
--loginopens a browser so you can solve it manually - If your session expires, re-run
uvx handshake-mcp-server --login
Cloudflare detection:
- Handshake uses Cloudflare bot protection. On any desktop with a GUI, always use
--no-headless(the setup wizard does this automatically) - On headless Linux servers, use
--virtual-displayto avoid the headless fingerprint - See Cloudflare Bot Detection for details
Session issues:
- Browser profile is stored at
~/.handshake-mcp/profile/ - Run
uvx handshake-mcp-server --statusto check if your session is still valid - Run
uvx handshake-mcp-server --logoutto clear the profile and start fresh
Installation issues:
- Ensure uv is installed:
curl -LsSf https://astral.sh/uv/install.sh | sh - Install the Chromium browser:
uvx handshake-mcp-serverwill install it automatically on first run
</details>
<br/>
🐳 Docker Setup
Prerequisites: Docker installed and running. No other dependencies needed.
Docker runs the server headless inside a container using an Xvfb virtual display — no browser window on your machine. The browser profile (cookies/session) is stored in a named Docker volume (handshake-profile) and persists across container restarts.
1. One-time login
docker compose build
docker compose run --rm -p 6080:6080 handshake-mcp --vnc-login
Open http://localhost:6080/vnc.html in your browser and log into Handshake manually.
2. Start the server
docker compose up -d
The MCP server is now available at http://127.0.0.1:8000/mcp (streamable-http transport).
3. Configure your MCP client
{
"mcpServers": {
"handshake": {
"command": "docker",
"args": [
"run", "--rm", "-i",
"-v", "handshake-profile:/home/pwuser/.handshake-mcp",
"handshake-mcp-server",
"--transport", "stdio",
"--virtual-display"
]
}
}
}
Docker Setup Help
<details> <summary><b>🔧 Useful Commands</b></summary>
docker compose logs -f # tail server logs
docker compose run --rm handshake-mcp --status # check session validity
docker compose down # stop the server
docker rmi -f handshake-mcp-server && docker volume rm -f handshake-profile # full reset
</details>
<details> <summary><b>❗ Troubleshooting</b></summary>
Login issues:
- Open
http://localhost:6080/vnc.htmlwithin 5 minutes of starting--vnc-login - If the session expires, repeat the one-time login step
- Chromium may leave lock files (
SingletonLock) in the profile dir after a container kill — the server cleans these up automatically on next start
Docker issues:
- Check Docker is running:
docker ps - Rebuild after pulling new code:
docker compose build - Full reset:
docker rmi -f handshake-mcp-server && docker volume rm -f handshake-profile
</details>
<br/>
🐍 Local Setup (Develop & Contribute)
Contributions are welcome! See CONTRIBUTING.md for architecture guidelines and the PR checklist. Please open an issue first to discuss your change before submitting a PR.
Prerequisites: Git and uv installed.
# 1. Clone the repository
git clone https://github.com/sudhxnva/handshake-mcp-server
cd handshake-mcp-server
# 2. Install dependencies
uv sync --group dev
# 3. Install the Chromium browser
uv run patchright install chromium
# 4. Log in (opens a browser window)
uv run -m handshake_mcp_server --login --no-headless
# 5. Start the server
uv run -m handshake_mcp_server --no-headless
Local Setup Help
<details> <summary><b>🔧 Development Commands</b></summary>
uv run ruff check . # lint
uv run ruff check . --fix # lint + auto-fix
uv run ruff format . # format
uv run ty check # type check
uv run pytest # tests
uv run pytest --cov # tests with coverage
Claude Desktop config for local dev:
{
"mcpServers": {
"handshake": {
"command": "uv",
"args": ["--directory", "/path/to/handshake-mcp-server", "run", "-m", "handshake_mcp_server"]
}
}
}
</details>
<details> <summary><b>❗ Troubleshooting</b></summary>
Scraping issues:
- Use
--no-headlessto watch browser actions and debug scraping problems - Add
--log-level DEBUGfor verbose logging
Python/Patchright issues:
- Requires Python 3.12+:
python --version - Reinstall Patchright:
uv run patchright install chromium - Reinstall dependencies:
uv sync --reinstall
</details>
<br/>
Cloudflare Bot Detection
Patchright does not bypass Cloudflare's headless fingerprint when headless=True — that mode uses the chromium-headless-shell binary, which is trivially detected.
| Environment | Solution |
|---|---|
| Desktop with GUI (macOS, Windows, Linux) | Use --no-headless — opens a real browser window |
| Headless Linux server | Use --virtual-display — Chrome runs non-headless against an Xvfb virtual display |
| Docker | Default CMD already uses --virtual-display |
Cloudflare challenge pages are detected automatically and raise a RateLimitError if unresolved.
<br/>
Tool Return Format
All scraping tools return:
{
"url": "string",
"sections": { "section_name": "raw text content" },
"references": { "section_name": [{ "kind": "job", "url": "/jobs/123", "text": "..." }] },
"section_errors": { "section_name": { "error_type": "...", "error_message": "..." } },
"unknown_sections": ["invalid_section_name"]
}
get_job_details also returns a metadata key with structured fields:
{
"metadata": {
"id": "123", "title": "Software Engineer Intern", "company": "Acme Corp",
"salary": 3000, "salary_type": "hourly",
"work_type": "hybrid", "locations": ["San Francisco, CA"],
"job_type": "internship", "employment_type": "part_time",
"start_date": "2025-06-01", "end_date": "2025-08-31",
"deadline": "2025-03-01", "posted_at": "2025-01-15",
"work_auth_required": false, "accepts_opt": true, "accepts_cpt": false,
"will_sponsor": true, "apply_url": "https://..."
}
}
[!NOTE] Salary values from the GraphQL API are in cents — divide by 100 to get dollars.
search_jobs also returns jobs (card-level metadata list) and job_ids. search_employers returns employer_ids. search_events returns event_ids.
<br/>
Acknowledgements
Built with FastMCP and Patchright. Inspired by the architecture of linkedin-mcp-server.
Use in accordance with Handshake's Terms of Service. Web scraping may violate platform terms. This tool is intended for personal use only.
License
Apache-2.0 — see LICENSE.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.