APIs and SDKs

Call Fireworks from custom agents, background runners, services, HTTP APIs, and SDKs

Call Fireworks from custom agents, background jobs, or services with curl, the OpenAI SDK, or the Anthropic SDK. Send requests to https://api.fireworks.ai/inference and use the same model ID across each supported wire API.

Compatible APIs

Choose the endpoint your SDK or client already uses:

Wire API	Endpoint	Typical clients
Chat completions (OpenAI-style)	POST /v1/chat/completions	OpenAI SDK, most LLM gateways
Messages (Anthropic-style)	POST /v1/messages	Anthropic SDK
Responses (OpenAI-style)	POST /v1/responses	OpenAI Responses API clients

For client setup, see OpenAI compatibility, Anthropic compatibility, or the Responses API.

Call the API

Chat completions

Most direct API clients use chat completions:

curl https://api.fireworks.ai/inference/v1/chat/completions \
  -H "Authorization: Bearer $FIREWORKS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "firerouter/opus",
    "messages": [{"role": "user", "content": "Say pong in one word."}]
  }'

If your account has an Anthropic Provider Key, that request can use Claude Opus. Otherwise, add -H "x-anthropic-api-key: $ANTHROPIC_API_KEY" to the request. To use Astra, set "model": "firerouter/astra" and connect an OpenAI Provider Key or send x-openai-api-key.

Messages (Anthropic-style)

Anthropic Messages clients use the same model id on /v1/messages:

curl https://api.fireworks.ai/inference/v1/messages \
  -H "Authorization: Bearer $FIREWORKS_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "firerouter/opus",
    "max_tokens": 1024,
    "messages": [{"role": "user", "content": "Say pong in one word."}]
  }'

Responses (OpenAI-style)

OpenAI Responses clients send input to /v1/responses:

curl https://api.fireworks.ai/inference/v1/responses \
  -H "Authorization: Bearer $FIREWORKS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "firerouter/opus",
    "input": "Say pong in one word."
  }'

Router IDs and provider credentials work the same way across all three APIs. If the account has no Anthropic Provider Key, add -H "x-anthropic-api-key: $ANTHROPIC_API_KEY".

Read the serving model

For a router request, the response model field identifies the model that served the request, such as glm-5p3, claude-opus-5-5, or gpt-6-astra. It does not repeat the router ID you sent.

Easy prompts tend to stay on the open model. Harder prompts tend to use the family named in the router ID. Neither result is guaranteed for a single request. Continue sending the short router slug, such as firerouter/opus.

Credentials

Credential	How to provide it	When
Fireworks API key (fw_...)	Authorization bearer token or X-Fireworks-Api-Key header	Every request; bills open-model traffic to your Fireworks account
Anthropic	Provider Keys, or x-anthropic-api-key	Bare firerouter, or any route containing a Claude family alias or Anthropic model ID
OpenAI	Provider Keys, or x-openai-api-key	Any route containing Astra or another OpenAI model

A provider key in a request header applies only to that request. A key connected through Provider Keys is available at the account level, so clients and gateways do not need to send it with every call.

Fireworks also accepts Anthropic credentials in x-api-key or Authorization: Bearer. One Authorization header cannot carry both the Fireworks and Anthropic keys. If it carries the Anthropic key, send the Fireworks key in X-Fireworks-Api-Key. Prefer x-anthropic-api-key for new integrations.

For an OpenAI SDK pointed at Fireworks, api_key is the Fireworks key. Pass the OpenAI provider key as a header:

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["FIREWORKS_API_KEY"],
    base_url="https://api.fireworks.ai/inference/v1",
    default_headers={"x-openai-api-key": os.environ["OPENAI_API_KEY"]},
)

Drop default_headers when an OpenAI Provider Key is already connected.

Limitations

• Accounts with data residency enabled must use a residency-compatible pinned serverless model. Other model-router requests are rejected.
• Model routers are not available through Microsoft Foundry.

Errors

Response	Cause	Fix
401	Fireworks key is missing or invalid	Send a valid fw_... key with Authorization: Bearer or X-Fireworks-Api-Key
Request rejected for data residency	The requested route is not residency-compatible	Pin a residency-compatible serverless model
404 Model id not found	Unknown id, or the account cannot use this router	Check the model id and account access
400 no_credential	The closed model you named has no provider key	Connect Provider Keys, or send the provider header
Upstream provider authentication failure	The provider key is wrong or revoked	Check the Anthropic or OpenAI key; exact status and message come from that provider

Model routers and deployment routers

A firerouter/... id is a model router. Routers load-balance traffic across your own Fireworks deployments.

See also

• FireRouter
• Provider Keys
• Routing Preferences