MCP Servers
MCP (Model Context Protocol) servers allow AI agents to query live data directly from Radarboard and connected services. There are two MCP categories in this project:| Category | Location | What it is |
|---|---|---|
| Built-in Radarboard MCP | Radarboard MCP | The MCP endpoint exposed by the web app itself at /api/mcp |
| External MCP servers | opencode.json and app settings | Third-party or local MCP servers such as RevenueCat and OpenPanel |
Built-in Radarboard MCP
Use the built-in Radarboard MCP when you want an external client such as ChatGPT to connect directly to your deployed Radarboard instance. The URL pattern is:External MCP servers
Radarboard also works with external MCP servers for individual services. This repo currently documents:| Server | Service | Type | Description |
|---|---|---|---|
| RevenueCat MCP | RevenueCat | Remote (hosted) | Query revenue metrics and subscription data |
| OpenPanel MCP | OpenPanel | Local (npm package) | Query analytics data for specific sites |
How MCP works
MCP servers expose tools that AI agents can call. For example, an AI agent could:- Ask “What’s the current MRR?” and query RevenueCat
- Ask “How many visitors did the site get today?” and query OpenPanel
- Ask “What are my unread Radarboard notifications?” and query the built-in Radarboard MCP
Where external servers are configured
External MCP servers are configured inopencode.json at the project root. Each entry specifies:
- The server type (remote URL or local command)
- Environment variables for authentication
- Any server-specific options
opencode.json