Get started with Fabric Gateway

This topic walks through connecting your first LLM provider or MCP server and sending a test request through the gateway.

Fabric get started

Connect a provider

Go to Getting started and choose from one of the following categories:

  • LLMs — Select a displayed LLM (Anthropic or OpenAI) or browse the LLM providers catalog.
  • MCP servers — Select an MCP server (GitHub, New Relic, Linear, or Supabase) or browse the MCP servers catalog.

If you select the available provider, you can connect it directly. Proceed to Complete the setup. If you browse the catalog, you can select a provider and then click Use integration to configure and activate it.

You can also select Services from the sidebar, and click Create connection to open an Add resource form with four service types: LLM (provider catalog), MCP (Model Context Protocol), A2A (agent-to-agent), and HTTP (plain REST/OpenAPI). Connecting a provider requires three pieces of information:

  • Provider name (lowercase letters, digits, ., _, -; must start with a letter or digit).
  • Target URL.
  • Upstream auth requirement — An auth type (for example, Bearer) and scope (Org/tenant, each user, or user group), plus an optional identity binding.

Complete the setup

Once a provider is connected, complete the following steps to create a route and send a test request through the gateway:

  1. Create a connection.
  2. Add your API key or token.
  3. Create a route.
  4. Create a gateway key.
  5. Send a test request.

The test request is shown inline:

  • Connecting OpenAI produces a POST /openai/v1/chat/completions call.
  • Connecting Anthropic produces a POST /anthropic/v1/messages call.
  • Connecting an MCP server (for example, New Relic) uses the same five-step flow, but the final step sends a tools/list request.

Send a test request from Postman

Routes can also be exercised directly from a Postman workspace rather than the product UI. A request to the Anthropic route is sent as:

POST https://<gateway-subdomain>.fabricgateway.ai/anthropic/v1/messages
Headers:
X-Gateway-Key: <gateway key>
anthropic-version: 2023-06-01
Content-Type: application/json

The response comes back as a normal Anthropic Messages API payload (200 OK), including the model name, a message ID, and the assistant’s reply. It confirms that the gateway is transparently proxying to the upstream while enforcing the gateway key.

Next steps