For the complete documentation index, see llms.txt. Markdown versions of all docs pages are available by appending .md to any docs URL.

Route agentgateway LLM traffic to the Meta Model API.

Configure Meta as an LLM provider in agentgateway. The meta preset supports the Chat Completions, Messages, and Responses APIs.

Before you begin

You need agentgateway 1.6.0 or later and a Meta Model API key. Create a key in the Meta Model API dashboard.

Configuration

Use the meta preset to route requests to the Meta Model API. The preset supplies the provider URL and supported request formats.

1. Set your API key in the terminal where you will run agentgateway.

   export META_API_KEY='<your-api-key>'

2. Create a config.yaml file with the following configuration.

   cat > config.yaml <<'EOF'
   # yaml-language-server: $schema=https://agentgateway.dev/schema/config
   llm:
     models:
     - name: "*"
       provider: meta
       params:
         apiKey: "$META_API_KEY"
         # Optional: use this upstream model for every matching request.
         # model: muse-spark-1.3
         # Optional: override the default provider URL.
         # baseUrl: https://api.meta.ai/v1
   EOF

   Setting	Description
   name	The model name to match in incoming requests. Use * to accept any model name.
   provider	The provider preset. Set to meta.
   params.apiKey	Your Meta Model API key. Use $META_API_KEY to read the key from the environment.
   params.model	Optional. The upstream model to use for every matching request. Omit to use the model from the request.
   params.baseUrl	Optional. Overrides the provider URL. Defaults to https://api.meta.ai/v1.

3. Validate the configuration.

   agentgateway -f config.yaml --validate-only

   Example output:

   Configuration is valid!

4. Start agentgateway.

   agentgateway -f config.yaml

Example request

From another terminal, send a Chat Completions request to the gateway. This example uses muse-spark-1.3, as shown in the Meta quickstart. Use a model that your Meta account can access.

curl http://localhost:4000/v1/chat/completions \
  -H 'Content-Type: application/json' \
  -d '{
    "model": "muse-spark-1.3",
    "messages": [{"role": "user", "content": "Hello from Meta!"}]
  }'