upwork-mcp
A Model Context Protocol server that enables AI agents to search, analyze, and apply for Upwork jobs autonomously. It facilitates specialized job scoring, proposal submission, and integrated client messaging through a combination of browser automation and API access.
README
upwork-mcp
A Model Context Protocol (MCP) server that connects Claude to Upwork — enabling AI agents to search jobs, analyze opportunities, submit proposals, and manage client communications autonomously.
Works for any freelance niche: development, design, writing, marketing, automation, and more.
Features
| Tool | Description |
|---|---|
search_jobs |
Search Upwork jobs by keyword, type, budget, experience level |
get_job_details |
Fetch full job description, screening questions, and client history |
analyze_job |
Score a job 0–100 across niche fit, client quality, budget, competition, and clarity |
submit_proposal |
Automatically write and submit a proposal with custom cover letter and bid |
get_proposals |
List submitted proposals and their current status |
get_messages |
Read conversations and check for unread messages |
send_message |
Reply to clients in ongoing conversations |
get_profile |
View your freelancer profile, JSS score, and connects balance |
update_profile |
Update your freelancer profile title, overview, and skills |
manual_login |
Capture session cookies from an already-logged-in Chrome tab |
Architecture
Claude Agent
│
▼
MCP Gateway (stdio, never restarts)
│ HTTP POST /tool
▼
Worker Server (port 47821, hot-reloads on code changes)
│
▼
CDP Proxy (port 9223, rewrites Host headers)
│
▼
Chrome on host (port 9222, Playwright CDP connection)
│
▼
Upwork Web UI
The gateway/worker split means you can update tool logic without restarting Claude or your MCP client.
Prerequisites
- Node.js 20+
- Docker + Docker Compose
- Google Chrome installed
- An Upwork freelancer account
Quick Start (Docker)
1. Clone and configure
git clone https://github.com/zcrossoverz/upwork-mcp.git
cd upwork-mcp
cp .env.example .env
Edit .env:
UPWORK_EMAIL=your@email.com
UPWORK_PASSWORD=yourpassword
FREELANCER_NAME=Your Name
FREELANCER_TITLE=Your Professional Title
FREELANCER_NICHE=your,skills,here
BID_RATE_DEFAULT=40
BID_RATE_MIN=25
BID_RATE_MAX=100
2. Start Chrome with CDP
connect-chrome-docker.bat
This launches Chrome with remote debugging + a CDP proxy that allows Docker to connect.
3. Start the worker
docker compose up -d
4. Capture your session
In Claude, call manual_login — it extracts cookies from your logged-in Chrome tab and saves the session. You only need to do this once (or after your session expires).
Connect to Claude Code
Add to your MCP config (~/.claude/claude_desktop_config.json):
{
"mcpServers": {
"upwork": {
"command": "node",
"args": ["/absolute/path/to/upwork-mcp/dist/gateway.js"],
"env": {
"WORKER_PORT": "47821"
}
}
}
}
Or via CLI:
claude mcp add upwork node /absolute/path/to/upwork-mcp/dist/gateway.js
Local Development (Hot Reload)
npm install
npm run worker # starts tsc --watch + nodemon on dist/worker.js
The gateway (dist/gateway.js) runs separately and proxies to the local worker — changes to any tool file reload the worker instantly without restarting Claude.
Recommended Agent Workflow
1. manual_login → Capture session (first time / session expired)
2. get_profile → Check connects balance before bidding
3. search_jobs → Find relevant jobs by keyword
4. get_job_details → Get full description + screening questions
5. analyze_job → Score the opportunity (skip grade D or F)
6. submit_proposal → Auto-bid with personalized cover letter
7. get_messages → Check for client responses
8. send_message → Reply to clients
9. get_proposals → Track active proposals
Example agent prompt
Search for freelance jobs matching my skills posted in the last 3 days.
For each job with grade A or B:
1. Get full job details
2. Analyze the opportunity
3. Write a personalized proposal highlighting my relevant experience
4. Submit at the recommended bid rate
Job Analysis Scoring
The analyze_job tool scores each job across 5 dimensions:
| Dimension | Max Points | What It Measures |
|---|---|---|
| Niche Fit | 30 | Keyword match between job and your FREELANCER_NICHE |
| Client Quality | 25 | Rating, total spend, hire rate, payment verified |
| Budget Fit | 20 | Budget vs your target rate, estimated project value |
| Competition | 10 | Number of existing proposals (fewer = better) |
| Project Clarity | 10 | Description detail, skills listed, budget specified |
| Red Flags | -5 each | Vague scope, low-budget signals, no client history |
Grades:
| Grade | Score | Action |
|---|---|---|
| A+ | 90–100 | APPLY NOW — high priority |
| A | 75–89 | APPLY — strong opportunity |
| B | 60–74 | APPLY — worth pursuing |
| C | 45–59 | CONSIDER — marginal fit |
| D | 30–44 | SKIP |
| F | <30 | AVOID |
Project Structure
upwork-mcp/
├── src/
│ ├── gateway.ts # MCP stdio gateway (thin proxy, never restarts)
│ ├── worker.ts # HTTP tool server (hot-reloads via nodemon)
│ ├── config.ts # Environment configuration
│ ├── browser/
│ │ ├── browser-manager.ts # Playwright CDP connection manager
│ │ └── upwork-auth.ts # Session management
│ └── tools/
│ ├── search-jobs.ts
│ ├── get-job-details.ts
│ ├── analyze-job.ts # Scoring engine
│ ├── submit-proposal.ts
│ ├── get-proposals.ts
│ ├── get-messages.ts
│ ├── send-message.ts
│ ├── get-profile.ts
│ ├── update-profile.ts
│ └── manual-login.ts # CDP cookie extractor
├── cdp-proxy.cjs # Host-side proxy: Docker → Chrome (fixes Host header)
├── connect-chrome-docker.bat # Launch Chrome + CDP proxy (for Docker use)
├── connect-chrome.bat # Launch Chrome only (for local use)
├── Dockerfile
├── docker-compose.yml
└── .env.example
Notes
Session Management
Login is handled manually via manual_login. Open Chrome, log in to Upwork normally (including any 2FA), then call manual_login — it captures all cookies and saves them for Playwright to reuse.
Bot Detection
Upwork may detect unusual activity. Use reasonable request intervals and do not run the agent continuously. The server connects to your real Chrome profile via CDP, which behaves more like a real browser than headless automation.
Selector Stability
Upwork's UI changes periodically. Browser-based selectors in src/tools/ may need updating if the UI is redesigned.
Ethical Use
- Only automate your own account
- Comply with Upwork's Terms of Service
- Review proposals before submission in production use
License
MIT
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.
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.
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.
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.