MCP Server
Connect AI assistants to Kaiten via the MCP (Model Context Protocol) server.
MCP Server
Kaiten includes an MCP (Model Context Protocol) server that lets AI assistants like Cursor, Claude Desktop, or other MCP clients interact with your Kaiten instance.
How It Works
The MCP server is a Python FastMCP application that auto-generates tools from the OpenAPI specification. AI assistants can then:
- List and create customers
- Manage instances and licenses
- Evaluate feature flags
- Check entitlement usage
- And any other API operation
Setup
1. Start with MCP Enabled
make up-fullThis starts the MCP server alongside the API.
2. Generate an API Token
Create a service account and token with the scopes you want the AI assistant to access:
curl -X POST http://localhost:6000/api/service-accounts \
-H "Authorization: Bearer <jwt>" \
-d '{"name": "MCP Assistant"}'
curl -X POST http://localhost:6000/api/service-accounts/<sa-id>/tokens \
-H "Authorization: Bearer <jwt>" \
-d '{"name": "mcp", "scopes": ["read:*", "write:*"]}'3. Configure Your MCP Client
Cursor
Add to your .cursor/mcp.json:
{
"mcpServers": {
"kaiten": {
"url": "http://localhost:6000/mcp",
"headers": {
"Authorization": "Bearer ksh_your_token_here"
}
}
}
}Claude Desktop
Add to your claude_desktop_config.json:
{
"mcpServers": {
"kaiten": {
"url": "http://localhost:6000/mcp",
"headers": {
"Authorization": "Bearer ksh_your_token_here"
}
}
}
}Health Check
Verify the MCP server is running:
curl http://localhost:6000/mcp/healthSecurity
The MCP server has full API access based on the token's scopes. Use restricted scopes for production AI assistants. The read:* scope is recommended for read-only access.