MCP Server
Connect UpHunt to Claude Desktop, Cursor, Cline, and Windsurf with the Model Context Protocol.
The UpHunt MCP server brings your AI-scored Upwork job feed into any Model Context Protocol client — Claude Desktop, Cursor, Cline, Windsurf, and Claude Code. Ask your AI tool to find the best-matched jobs and research clients in plain language, without leaving your editor.
It is a thin, read-only bridge over the UpHunt API. For programmatic auto-apply, AI cover letters, and webhooks, use the full UpHunt API.
The MCP server runs locally on your machine and authenticates with your UpHunt API key. It never logs into Upwork or scrapes the site — it only reads data from UpHunt's own API, so your Upwork account is never touched.
Setup
Get your API key
Open your Dashboard, go to Auto-Apply → API & Webhooks, and click Generate API Key. Copy it — you'll only see it once. See Authentication for details.
Add UpHunt to your MCP client
Add the server to your client's MCP config and paste your key:
The same block works in Cursor (mcp.json), Cline, and Windsurf — they all read the standard mcpServers shape.
Restart and ask
Restart your client. Try:
Find React jobs over $5,000 with a match score above 8 and summarize the top client.
Tools
The server exposes four read-only tools over your UpHunt data:
| Tool | What it does |
|---|---|
search_jobs | Search your live, AI-scored Upwork job feed by keyword, budget, job type, and minimum match score. |
get_job | Pull the full details of any Upwork job by URL or ciphertext, including budget, skills, and screening questions. |
get_client | Read a buyer's open jobs, hire history, and spend stats before you bid. |
list_applied_jobs | Review the jobs you've already applied to, with their match scores and status. |
The MCP server does not submit proposals. Reading and drafting happen in your AI tool; sending a proposal is a deliberate step you take through the API or dashboard, where a human stays in control.
From research to applying
A common workflow pairs the MCP server with the API:
Research in your AI tool
Use the MCP server to surface high-scoring jobs and vet clients conversationally.
Apply via the API
Submit proposals programmatically with the same API key.
Let AI write the proposal
Generate tailored cover letters with /generate-proposal.
Automate via webhooks
React to every matched job in real time.
Troubleshooting
Server not showing up? Make sure npx is on your PATH and fully restart the client (quit, don't just reload). Check that UPHUNT_API_KEY is set and valid — an invalid key returns 401, same as the REST API. See Errors.