Skip to content

Choose an Integration

Connect the client you already use to Mission Control. Model APIs accept ordinary user or service virtual keys; browser sign-in creates a client connection for MCP tools. Each request uses the current access of its user or service account in one Team.

If this is your first request, follow Make your first model request or Call your first MCP tool.

Find your integration

Client What the guide helps you do Credential
OpenAI Python and TypeScript Call models, stream a response, and use the supported Responses subset User or service virtual key
Claude Code Add remote MCP tools and sign in from your browser MCP client connection, or user/service virtual key
MCP TypeScript client Discover tools, call them, and handle access, connection, and approval results User or service virtual key; browser authorization for interactive clients
PydanticAI Run your own agent with governed models, tools, and telemetry User/service virtual key with telemetry.ingest
OpenTelemetry Export your application's spans and connect them to gateway requests Virtual key with telemetry.ingest

Prepare your connection

Get the gateway URL and Console URL from your deployment administrator. Export credentials in the process that runs your client. Example URLs such as https://gateway.example.test and keys such as mck_example_not_a_real_key are fictional; replace them before running a sample.

Use a model returned by GET /v1/models and a tool returned by tools/list. Availability depends on Team allowance, the key's current operation/resource policy, and the operation's limits. A provider name or provider credential is not a Mission Control model identifier or key.

Compatibility and next steps