MCP integration for IDEs
Senix integrates with any MCP-compatible IDE. Use it in Cursor, Claude Code, Windsurf, and others.
What is MCP?
The Model Context Protocol (MCP) is an open standard for AI tools to communicate with external services. IDEs that support MCP can connect to a server like Senix and expose its tools to your AI assistant. When you ask your IDE's AI to review your changes, it calls the Senix tool, sends the diff, and gets back our analysis — no pull request required.
Fastest setup: the dashboard
The quickest path is Dashboard → Connect IDE. It generates a token, gives you a one-click Add to Cursor button (or a copyable claude mcp add command for Claude Code), and a live Test connection button. The steps below are the manual equivalent.
Manual setup
- Go to Dashboard → MCP tokens and click Generate token.
- Copy the token — it is shown once (with a 60-second recovery window).
- Add it to your IDE using the per-IDE instructions below, replacing
YOUR_TOKEN_HEREwith the copied token.
Cursor
Use the one-click Add to Cursor button on the Connect IDE page, or add this to your ~/.cursor/mcp.json (global) or .cursor/mcp.json (per project).
{
"mcpServers": {
"senix": {
"transport": "http",
"url": "https://senix-chi.vercel.app/api/mcp",
"headers": {
"Authorization": "Bearer YOUR_TOKEN_HERE"
}
}
}
}Claude Code
Run this in your terminal, then restart Claude Code:
claude mcp add senix --transport http \ --url https://senix-chi.vercel.app/api/mcp \ --header "Authorization: Bearer YOUR_TOKEN_HERE"
Or add it to your .mcp.json manually:
{
"mcpServers": {
"senix": {
"type": "http",
"url": "https://senix-chi.vercel.app/api/mcp",
"headers": {
"Authorization": "Bearer YOUR_TOKEN_HERE"
}
}
}
}Windsurf
Add the following to your Windsurf MCP config at ~/.codeium/windsurf/mcp_config.json.
{
"mcpServers": {
"senix": {
"serverUrl": "https://senix-chi.vercel.app/api/mcp",
"headers": {
"Authorization": "Bearer YOUR_TOKEN_HERE"
}
}
}
}How to use it in your IDE
Once connected, ask your IDE's AI in plain language:
- "Review my changes with Senix"
- "Check this code for risks"
The AI calls Senix's review_changestool with your before/after file contents. You'll see the behavioral summary, risk level, and focus areas right in your chat panel.
Differences from the GitHub bot
- On-demand vs. automatic — MCP runs when you ask; the GitHub bot runs automatically on PR open and push.
- Uncommitted vs. committed — MCP works on uncommitted changes in your editor; the GitHub bot needs a pull request.
- Same brain — both use the same prompt and the same 8-flag risk taxonomy, so the analysis is consistent across surfaces.
Need help generating a token? See Configuration for managing MCP tokens.
