Fireworks Provider Setup: Auth and Model Selection

Fireworks provides open-weight and routed models through an API that works like OpenAI's. To access the current Fire Pass GLM router, two pre-cataloged Kimi models, and any Fireworks model or router id at runtime, install the official Fireworks provider plugin.

Getting started

Install the plugin

openclaw plugins install @openclaw/fireworks-provider

Set the Fireworks API key

openclaw onboard --auth-choice fireworks-api-key

openclaw onboard --non-interactive --accept-risk --skip-health \
  --auth-choice fireworks-api-key \
  --fireworks-api-key "$FIREWORKS_API_KEY"

export FIREWORKS_API_KEY=fw-...

When you onboard, the key is saved under the fireworks provider in your auth profiles, and Fireworks' current Fire Pass GLM 5.2 Fast router gets set as the default model.

Verify the model is available

openclaw models list --provider fireworks

Make sure GLM 5.2 Fast, Kimi K2.6, and Kimi K2.6 Fast appear in the list. If FIREWORKS_API_KEY cannot be resolved, openclaw models status --json flags the missing credential under auth.unusableProfiles.

Non-interactive setup

For automated or CI setups, supply all parameters directly on the command line:

openclaw onboard --non-interactive \
  --mode local \
  --auth-choice fireworks-api-key \
  --fireworks-api-key "$FIREWORKS_API_KEY" \
  --skip-health \
  --accept-risk

Built-in catalog

Note

OpenClaw locks every Fireworks Kimi model to thinking: off because, unless the request explicitly turns off thinking, Kimi on Fireworks can expose chain-of-thought in the visible reply. If you route the same model through Moonshot directly, Kimi reasoning output is preserved. To switch between providers, check thinking modes.

Custom Fireworks model ids

At runtime, OpenClaw will take any Fireworks model or router id. Use the exact id shown by Fireworks and prepend fireworks/ to it. Dynamic resolution relies on the Fire Pass template's OpenAI-compatible API and tags GLM ids as text-only; other dynamic ids claim text + image input. When the id matches the Kimi pattern, thinking is turned off automatically. For a model with different capabilities, set up a custom model entry that lists its supported input types.

{
  agents: {
    defaults: {
      model: {
        primary: "fireworks/accounts/fireworks/models/<your-model-id>",
      },
    },
  },
}

How model id prefixing works

In OpenClaw, every Fireworks model ref begins with fireworks/, followed by the exact id or router path from the Fireworks platform. For instance:

• Router model: fireworks/accounts/fireworks/routers/kimi-k2p6-turbo
• Direct model: fireworks/accounts/fireworks/models/<model-name>

When building the API request, OpenClaw removes the fireworks/ prefix and sends the remaining path to the Fireworks endpoint as the OpenAI-compatible model field.

Why thinking is forced off for Kimi

Since Fireworks delivers Kimi without a separate reasoning channel, chain-of-thought may show up in the visible content stream. For every Fireworks Kimi request, OpenClaw sends thinking: { type: "disabled" } and removes reasoning, reasoning_effort, and reasoningEffort from the payload (extensions/fireworks/stream.ts). The provider policy (extensions/fireworks/thinking-policy.ts) advertises only the off thinking level for Kimi model ids, so manual /think switches and provider-policy surfaces match the runtime contract.

To get Kimi reasoning end-to-end, set up the Moonshot provider and route the same model through it.

Environment availability for the daemon

If the Gateway runs as a managed service (launchd, systemd, Docker), the Fireworks key must be accessible to that process, not just to your interactive shell.

Warning

A key exported only in an interactive shell won't help a launchd or systemd daemon unless that environment is also imported there. To make the key readable from the gateway process, set it in ~/.openclaw/.env or via env.shellEnv.

OpenClaw reads ~/.openclaw/.env at config load time, so any keys placed there are available to managed gateway services across all platforms. After you rotate the key, restart the gateway or execute openclaw doctor --fix again.

Related

• Model providers, How to pick providers, reference models, and set up failover.

• Thinking modes, /think settings, provider rules, and directing reasoning-capable models.

• Moonshot, Use Kimi with native thinking output via Moonshot's API.

• Troubleshooting, Common issues and answers.