Claude Code

no
Summary: Use Claude Code with Fireworks AI models

Original Documentation

Documentation Index#

Fetch the complete documentation index at: https://docs.fireworks.ai/llms.txt Use this file to discover all available pages before exploring further.

Use Claude Code with Fireworks AI models

Fireworks offers 200+ models—set the model endpoint in your configuration; each model lists the value on its page in the Fireworks Model Library.

Fireworks AI supports Claude Code through our Anthropic-compatible API, enabling you to use powerful open-source models as a drop-in replacement for Claude’s models.

FireConnect routes Claude Code through Fireworks. Install the Claude Code CLI first.

    curl -fsSL https://raw.githubusercontent.com/fw-ai/fireconnect/main/install.sh | bash
    ```

```bash
    curl -fsSL https://raw.githubusercontent.com/fw-ai/fireconnect/main/install.sh | FIREWORKS_API_KEY="fpk_..." bash
    ```


The installer walks you through three steps in your terminal:

1. **API key** — paste a [Fireworks API key](https://app.fireworks.ai/settings/users/api-keys). If you use [Fire Pass](/firepass), paste your dedicated Fire Pass key (`fpk_...`), not a standard `fw_...` key.
2. **Plugin** — adds the FireConnect plugin to Claude Code so you can use `/fireconnect:*` commands.
3. **Configuration** — saves `~/.claude/settings.json` so Claude Code sends model requests to Fireworks instead of Anthropic.
<span class="step-end"></span>

<span class="step-marker" data-step-title="Restart Claude Code"></span>
Restart Claude Code.
<span class="step-end"></span>
<span class="steps-end"></span>

| Slot        | Model ID                                     |
| ----------- | -------------------------------------------- |
| Opus (main) | `accounts/fireworks/routers/kimi-k2p6-turbo` |
| Sonnet      | `accounts/fireworks/models/glm-5p1`          |
| Haiku       | `accounts/fireworks/models/minimax-m2p5`     |
| Subagent    | `accounts/fireworks/models/minimax-m2p5`     |

Use `/fireconnect:set-models` to change models, or `/fireconnect:off` and `/fireconnect:on` to toggle routing.

### Using Fire Pass

[Fire Pass](/firepass) covers the default Opus model (`kimi-k2p6-turbo`). Use a dedicated **Fire Pass API key** (`fpk_...`) from [app.fireworks.ai/api-keys](https://app.fireworks.ai/api-keys)—not a standard Fireworks key (`fw_...`).

To use the Fire Pass model for every slot:

```text
/fireconnect:set-models --opus kimi-k2p6-turbo --sonnet kimi-k2p6-turbo --haiku kimi-k2p6-turbo --subagent kimi-k2p6-turbo
Install without the shell script:
/plugin marketplace add https://github.com/fw-ai/fireconnect.git
/plugin install fireconnect@fireconnect

If slash commands are unavailable:

fireconnect status
fireconnect on --api-key fpk_...
fireconnect off
fireconnect set --main kimi-k2p6-turbo --haiku minimax-m2p5

Manual configuration#

Skip if you used FireConnect. Pick a model tab below for settings.json or environment-variable examples.

Moonshot’s flagship agent model for coding, reasoning, and long-context tasks (256k context).

settings.json#

Create ~/.claude/settings.json (or ./.claude/settings.json in a project):

    {
        "$schema": "https://json.schemastore.org/claude-code-settings.json",
        "apiKeyHelper": "bash -c 'echo your-fireworks-api-key'",
        "env": {
            "ANTHROPIC_BASE_URL": "https://api.fireworks.ai/inference",
            "ANTHROPIC_MODEL": "accounts/fireworks/models/kimi-k2p6",
            "ANTHROPIC_SMALL_FAST_MODEL": "accounts/fireworks/models/kimi-k2p6",
            "ANTHROPIC_DEFAULT_SONNET_MODEL": "accounts/fireworks/models/kimi-k2p6",
            "ANTHROPIC_DEFAULT_HAIKU_MODEL": "accounts/fireworks/models/kimi-k2p6",
            "ANTHROPIC_DEFAULT_OPUS_MODEL": "accounts/fireworks/models/kimi-k2p6"
        },
        "model": "accounts/fireworks/models/kimi-k2p6"
    }
    ```

### Environment variables

```bash
    export ANTHROPIC_BASE_URL="https://api.fireworks.ai/inference"
    export ANTHROPIC_AUTH_TOKEN="your-fireworks-api-key"
    export ANTHROPIC_MODEL="accounts/fireworks/models/kimi-k2p6"
    export ANTHROPIC_SMALL_FAST_MODEL="accounts/fireworks/models/kimi-k2p6"
    export ANTHROPIC_DEFAULT_SONNET_MODEL="accounts/fireworks/models/kimi-k2p6"
    export ANTHROPIC_DEFAULT_HAIKU_MODEL="accounts/fireworks/models/kimi-k2p6"
    export ANTHROPIC_DEFAULT_OPUS_MODEL="accounts/fireworks/models/kimi-k2p6"
    ```
  <span class="tab-end"></span>

  <span class="tab-start" data-tab-title="GLM-5.1"></span>
Zhipu AI's model for code generation, agents, and long-context tasks.

### settings.json

Create `~/.claude/settings.json` (or `./.claude/settings.json` in a project):

```json
    {
        "$schema": "https://json.schemastore.org/claude-code-settings.json",
        "apiKeyHelper": "bash -c 'echo your-fireworks-api-key'",
        "env": {
            "ANTHROPIC_BASE_URL": "https://api.fireworks.ai/inference",
            "ANTHROPIC_MODEL": "accounts/fireworks/models/glm-5p1",
            "ANTHROPIC_SMALL_FAST_MODEL": "accounts/fireworks/models/glm-5p1",
            "ANTHROPIC_DEFAULT_SONNET_MODEL": "accounts/fireworks/models/glm-5p1",
            "ANTHROPIC_DEFAULT_HAIKU_MODEL": "accounts/fireworks/models/glm-5p1",
            "ANTHROPIC_DEFAULT_OPUS_MODEL": "accounts/fireworks/models/glm-5p1"
        },
        "model": "accounts/fireworks/models/glm-5p1"
    }
    ```

### Environment variables

```bash
    export ANTHROPIC_BASE_URL="https://api.fireworks.ai/inference"
    export ANTHROPIC_AUTH_TOKEN="your-fireworks-api-key"
    export ANTHROPIC_MODEL="accounts/fireworks/models/glm-5p1"
    export ANTHROPIC_SMALL_FAST_MODEL="accounts/fireworks/models/glm-5p1"
    export ANTHROPIC_DEFAULT_SONNET_MODEL="accounts/fireworks/models/glm-5p1"
    export ANTHROPIC_DEFAULT_HAIKU_MODEL="accounts/fireworks/models/glm-5p1"
    export ANTHROPIC_DEFAULT_OPUS_MODEL="accounts/fireworks/models/glm-5p1"
    ```
  <span class="tab-end"></span>
<span class="tab-group-end"></span>

<span class="callout-start" data-callout-type="tip"></span>
  Set all model override flags so subagents (Explore, Web Search, etc.) use Fireworks instead of Anthropic's Claude. With `settings.json`, `apiKeyHelper` handles auth—you do not need `ANTHROPIC_AUTH_TOKEN` in `env`.
<span class="callout-end"></span>

Run `claude` to start. Generate API keys at [app.fireworks.ai](https://app.fireworks.ai).

## Middleware/Proxy Providers

Use the same `settings.json` shape as [Manual configuration](#manual-configuration). Change only `ANTHROPIC_BASE_URL` (and Portkey headers where noted).

### LiteLLM Proxy (Self-Hosted)

```yaml
general_settings:
  allow_client_side_credentials: true
  pass_through_endpoints:
    - path: "/fw-anthropic"
      target: "https://api.fireworks.ai/inference"
      include_subpath: true
      forward_headers: true

Set ANTHROPIC_BASE_URL to https://<YOUR_LITELLM_HOST>/fw-anthropic.

Portkey#

Create an Anthropic provider in Portkey with custom host https://api.fireworks.ai/inference/v1 and your Fireworks API key.

Set ANTHROPIC_BASE_URL to https://api.portkey.ai, ANTHROPIC_AUTH_TOKEN to your Portkey key, and add headers:

x-portkey-api-key: your-portkey-api-key
x-portkey-provider: your-custom-fireworks-provider
anthropic-version: 2023-06-01

Why Use Fireworks with Claude Code?#

  • Cost Savings: Significantly lower cost per token compared to Claude’s native pricing
  • Adaptive serverless limits: Throughput scales with usage instead of relying on a fixed hourly subscription tier. See Serverless rate limits.
  • Model Choice: Access to the latest open-source models
  • Privacy: Your code stays within your chosen infrastructure
  • Transparent Pricing: Clear per-token pricing without subscription tiers

Next Steps#

Need Help?#

For assistance with Claude Code integration, contact our team or join our Discord community.

Link last verified June 7, 2026. View original ↗
Source: Fireworks AI Docs
Link last verified: 2026-06-07