Authentication

FireRouter uses a bring-your-own-key contract. The service validates your Fireworks key on every request and forwards your provider key only on pass-through legs to closed-source models. FireRouter does not store provider keys server-side.

Required keys

Key	Header or env	Used for
Fireworks API key (`fw_...`)	`Authorization: Bearer` or `X-Fireworks-Api-Key`	FireRouter auth and redirected Fireworks inference
Anthropic API key (`sk-ant-...`)	`x-anthropic-api-key`	Calls to Claude models
OpenAI API key (`sk-...`)	`x-openai-api-key`	Calls to OpenAI models

The default firerouter model uses Claude Opus 5 as its primary model, so every request requires an Anthropic key. FireRouter fails closed when the primary model’s credential is missing; it does not silently restrict the route to Fireworks models. For a model-specific FireRouter slug, the first model’s credential is required. Later models are eligible only when their credentials are present. Slugs that contain only Fireworks models need no additional provider key beyond the Fireworks API key.

Fireworks key header

Send your Fireworks API key with either header:

-H "Authorization: Bearer $FIREWORKS_API_KEY"
# or
-H "X-Fireworks-Api-Key: $FIREWORKS_API_KEY"

X-FireRouter-Fireworks-Key is also accepted. New integrations should prefer X-Fireworks-Api-Key.

Anthropic provider key

The canonical header for Anthropic pass-through is:

-H "x-anthropic-api-key: $ANTHROPIC_API_KEY"

x-api-key and Authorization: Bearer are also accepted for Anthropic clients that send credentials that way. For new integrations, prefer x-anthropic-api-key. Example:

curl https://api.fireworks.ai/inference/v1/chat/completions \
  -H "Authorization: Bearer $FIREWORKS_API_KEY" \
  -H "x-anthropic-api-key: $ANTHROPIC_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "firerouter", "messages": [...]}'

OpenAI provider key

When the selected FireRouter slug includes an OpenAI model, send your OpenAI API key as:

-H "x-openai-api-key: $OPENAI_API_KEY"

Do not put the OpenAI provider key in the OpenAI client’s api_key field. That field supplies the Authorization header used to authenticate to the Fireworks gateway, so it must contain your Fireworks API key. Pass the OpenAI provider key as a custom 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"]},
)

Common errors

Response	Cause	Fix
`401 Missing Fireworks API key`	Fireworks key header missing or empty	Set `X-Fireworks-Api-Key` to your `fw_...` key
`401 invalid Fireworks API key`	Key rejected	Confirm the key is valid in the dashboard
`400` with `no_credential`	Required provider key missing or sent under the wrong header	Send the key under the provider-specific header
Provider `401`	Provider key is invalid	Check the key sent as `x-anthropic-api-key` or `x-openai-api-key`

Quickstart: API call examples
Claude Code (manual setup): settings.json setup
Overview: routing model and model ID

⌘I

Get Started

Serverless Inference

Deployments

Models & Inference

Training

Managed Fine-Tuning

Training API

Fire Pass

Fireworks Nexus

Administration

Security & Compliance

Integrations

Reference

Required keys

Fireworks key header

Anthropic provider key

OpenAI provider key

Common errors

​Required keys

​Fireworks key header

​Anthropic provider key

​OpenAI provider key

​Common errors

​Related

Required keys

Fireworks key header

Anthropic provider key

OpenAI provider key

Common errors

Related