The Paragon MCP server gives your AI coding tools access to code review, testing, and analysis capabilities. Connect it to Claude Code, Cursor, or Windsurf to use Paragon directly from your development environment.
Prerequisites
Get your API key
- Go to app.paragon.run
- Sign in or create an account
- Navigate to Settings → API Keys
- Copy your API key
Some tools like generate_tests, run_paragon, list_test_suites, and save_to_suite require an API key. Other tools work without one.
Claude Code
Cursor
Windsurf
claude mcp add-json paragon '{"command":"paragon","args":["mcp-server"],"env":{"POLARITY_API_KEY":"your-api-key-here"}}'
your-api-key-here with your API key. Verify with:Full Claude Code setup guide →Open Settings → MCP → + Add new MCP server, or add to .cursor/mcp.json:{
"mcpServers": {
"paragon": {
"command": "paragon",
"args": ["mcp-server"],
"env": {
"POLARITY_API_KEY": "your-api-key-here"
}
}
}
}
your-api-key-here with your API key. A green status indicator confirms the connection.Full Cursor setup guide → Open Cascade → Plugins (MCP) → Add custom server, or add to ~/.codeium/windsurf/mcp_config.json:{
"mcpServers": {
"paragon": {
"command": "paragon",
"args": ["mcp-server"],
"env": {
"POLARITY_API_KEY": "your-api-key-here"
}
}
}
}
your-api-key-here with your API key. Restart Windsurf and check for a green status indicator. Code review
| Tool | Description |
|---|
run_paragon | Run a Paragon AI prompt — reads/writes files, runs shell commands, and performs complex coding tasks. Requires API key. |
list_reviewed_prs | List pull requests with Paragon review comments in the current repository. Supports filtering by state (open, closed, all). |
get_review_comments | Get parsed review findings from a PR, including severity, descriptions, file locations, and suggested fixes. Auto-detects PR from current branch if omitted. |
resolve_review_comment | Reply to a review comment and optionally resolve the thread. |
Testing
| Tool | Description |
|---|
detect_test_framework | Scan the repository and detect test frameworks in use. Returns framework details, run commands, file patterns, and confidence level. |
find_tests | Discover existing test files using framework-aware file patterns. Returns file paths and count. |
run_tests | Execute unit/integration tests using the detected or specified framework. Returns pass/fail status and full output. |
run_e2e_tests | Run Playwright E2E tests locally. Installs browsers if needed, optionally starts a dev server, and captures results. |
generate_tests | Generate unit, integration, or E2E tests using Paragon AI. Detects patterns, writes tests, verifies compilation, runs them, and fixes failures. Requires API key. |
Test suite management
| Tool | Description |
|---|
list_test_suites | List your test repositories and suites from the Paragon cloud dashboard. Requires API key. |
save_to_suite | Save locally generated test files to a Paragon cloud test suite. Requires API key. |
Utility
| Tool | Description |
|---|
check_setup | Check if the MCP server is properly configured. Reports the status of the API key, Paragon binary, and GitHub CLI. |
Example prompts
After setup, your AI tool can automatically use Paragon when relevant. You can also request tools directly:
Review the comments on my current PR and fix the issues
Generate unit tests for src/auth.ts
Run the test suite and show me what's failing
Find all Playwright tests and run them in headed mode
