Model Context Protocols (MCPs)
MCP (Model Context Protocol) allows AI-powered editors to connect AI directly to your services—whether it’s GitHub for pull requests, Supabase for databases, or PostHog for analytics.
This turns the LLM from a passive assistant into an active agent that can interact with APIs, services, or CLIs.
Built-In MCPs
Zap.ts comes preconfigured with the most commonly used services to save you setup time. These are defined in configuration files for each editor:
- Cursor:
.cursor/mcp.json
- VS Code:
.vscode/mcp.json
- Windsurf:
.windsurf/mcp.json
Here are some of the built-in MCPs:
- ElevenLabs — Use realistic text-to-speech directly from your IDE.
- Firecrawl — Crawl and extract structured data from websites.
- GitHub — Automate PRs, issues, and more.
- Magic by 21st.dev — Secure authentication with a few lines of code.
- Neon — A modern Postgres platform, optimized for serverless.
- Notion — Read and write structured Notion content via the API.
- Perplexity — Ask factual questions via API.
- PostHog — Analyze events, user sessions, and feature flags.
- Resend — Send transactional emails with ease.
- Sentry — Monitor and manage application errors from inside Cursor.
- Supabase — Query and mutate your database directly.
Of course, you can customize them as you wish.
Environment Setup
Many of these integrations rely on environment variables for API keys or tokens.
All MCP-related variables follow a naming convention: they are prefixed with MCP_
.
Make sure to define them in your local .env
file or shell environment.
Example:
MCP_POSTHOG_AUTH_HEADER=Bearer your_posthog_token
MCP_SUPABASE_ACCESS_TOKEN=your_supabase_token