Use the gateway
These topics are for people who call the gateway: authenticating, finding what’s available, sending requests to LLMs, MCP servers, and REST APIs, and fixing problems.
Find the gateway’s base URL, send your gateway key in the fg-api-key header, and make your first request through Fabric AI Gateway.
Find out which models, MCP servers, and APIs your gateway exposes, which path reaches each one, and whether it needs a gateway key.
Call LLM APIs
Name a model in your request, list the models a route serves, and understand the errors for a model the gateway can’t serve.
Call chat completions, responses, embeddings, and the model list through an LLM route, with streaming and OpenAI-shaped errors.
Call messages, token counting, and the model list through an LLM route, with streaming and Anthropic-shaped errors.
Connect to MCP servers
Point an MCP client at an MCP route, authenticate with a gateway key or by signing in, and check that the client sees the route’s tools.
Connect your own account on an upstream service, such as GitHub or Atlassian, so the gateway can call that service’s MCP server for you.
Call REST APIs
Troubleshoot
Work out which step refused your request — the gateway key, access control, or the upstream credential — and who can fix it.
Recognize gateway errors on REST, LLM, and MCP routes, decide which ones to retry, and read rate-limit responses and timeouts.

