MCP (AI Assistant Integration) on GekkoVet Developers

Connecting an AI Assistant

Mon, 01 Jan 0001 00:00:00 +0000

Connecting an AI Assistant#

GekkoVet’s MCP server can be connected to any MCP-compatible client. The endpoint and transport are the same for all of them:

URL: https://mcp.gekkovet.workers.dev/mcp
Transport: Streamable-HTTP (recommended) or SSE at /sse
Auth: OAuth 2.0 — handled interactively on first connect

Below are setup instructions for the most common clients.

Claude Desktop#

Claude Desktop currently speaks MCP over local stdio only, so a small bridge is needed to connect it to the remote GekkoVet server. Use mcp-remote:

Authentication

Mon, 01 Jan 0001 00:00:00 +0000

MCP Authentication#

Unlike the GekkoVet REST API (which uses the client credentials grant), the MCP server authenticates end-users interactively using the OAuth 2.0 authorization code flow. This means every MCP tool call runs on behalf of a real GekkoVet user account — the same account you use to log into GekkoVet itself.

The AI assistant never sees your GekkoVet password. It only ever holds an opaque access token issued by the MCP server.

Tool Reference

Mon, 01 Jan 0001 00:00:00 +0000

Tool Reference#

Every tool exposed by the GekkoVet MCP server is listed here, grouped by domain. The AI assistant chooses tools automatically based on what you ask — this reference is primarily for understanding capabilities and debugging.

Tool definitions are generated from the same function registry that powers GekkoVet’s own assistants, so the canonical list always matches what’s running in production. To see the exact JSON schema for a given tool, invoke tools/list via the MCP Inspector.