Cursor Quick Start
Add ToolPilot to Cursor and start discovering developer tools in under 2 minutes.
Prerequisites
- Cursor editor installed
- Node.js 18+ (for
npx)
Step 1: Add the MCP Server Config
Create or edit the MCP configuration file in your project's .cursor/ directory:
.cursor/mcp.json
1{
2 "mcpServers": {
3 "toolpilot": {
4 "command": "npx",
5 "args": ["-y", "@anthropic/toolpilot-mcp"]
6 }
7 }
8}
Where to place the config
You can add this at the project level (
.cursor/mcp.json in your repo root) or the global level (~/.cursor/mcp.json). Project-level configs are recommended so team members share the same setup.Cursor Settings UI
You can also add MCP servers through Cursor Settings → Features → MCP Servers. Click “Add new MCP server”, choose “command” type, and enter
npx -y @anthropic/toolpilot-mcp as the command.Step 2: Try Your First Search
Open Cursor's AI chat (Cmd+L / Ctrl+L) and ask about developer tools. ToolPilot handles tool discovery through the MCP protocol:
Example conversation
You: "I need a TypeScript ORM that supports PostgreSQL, has good migration
tooling, and works well with serverless."
Cursor (using ToolPilot):
→ Calls search_tools with your query
→ ToolPilot may ask: "Do you need edge-runtime support or standard Node.js?"
→ Cursor answers via search_tools_respond
→ Returns ranked results:
1. Drizzle ORM — Health: Active ★★★★★ (lightweight, edge-ready)
2. Prisma — Health: Active ★★★★☆ (rich ecosystem, schema-first)
3. Kysely — Health: Active ★★★★☆ (type-safe query builder)
4. TypeORM — Health: Stable ★★★☆☆ (mature, decorator-based)Agent mode
Make sure you're using Cursor in Agent mode (not Ask or Edit mode) so it can call MCP tools automatically.
Step 3: Explore Results
Each recommendation from ToolPilot includes:
- Health tier — Active, Stable, Slowing, or At Risk
- Graph context — Alternatives, companions, and dependencies
- Key metadata — Stars, license, language, and last release
- Match rationale — Why this tool fits your specific requirements
Ask follow-ups like “Compare Drizzle and Prisma for my use case” or “What testing tools pair well with Drizzle?”.
Step 4: Report Outcomes
After adopting a tool, tell Cursor how it went. This feeds back into ToolPilot's graph and improves future results:
Feedback example
You: "Drizzle ORM is working perfectly for my serverless project."
Cursor (using ToolPilot):
→ Calls report_outcome with tool="drizzle-orm", outcome="adopted"
→ The graph learns from your feedback