MCP Reference

MCP reference

MCP (the Model Context Protocol) is the common language AI assistants use to connect to tools. Your camera speaks it natively, so assistants like Claude can see and record what your camera sees, all on your own network.

MCP server

Your camera has an MCP server built right in, available at http://eyeofthetiger.local/mcp/. Point an MCP-compatible assistant at that address and it can take photos and record clips for you.

💡
Keep the trailing slash. A request to /mcp without it gets redirected, and some clients will not follow that redirect.

Available tools

ToolDescriptionParameters
take_snapshotCapture a still image and return itfull (bool, default false) — pass true for a one-shot 12 MP photo instead of an instant stream grab
record_videoRecord a fixed-length clip and return a download URL for the MP4duration_s (integer, 1–60, default 10)
get_statusReturn the full camera activity picture: recording state and the current qualityNone
set_qualitySet the quality preset (stills and clips)level: high, medium, or low

record_video returns a download URL (http://eyeofthetiger.local/v1/library/{id}, pointing straight at the clip’s library entry) rather than the video itself, since an assistant cannot view video. Fetch the URL to save the MP4.


For step-by-step client setup, see the Integration Guides. To confirm the server is reachable, try the Quick Checks.