na{xx}en

OpenClaw

Configure OpenClaw to use naxxen as a proxy for all LLM providers.

OpenClaw supports custom provider base URLs. You can create naxxen-proxied providers alongside your direct ones.

How it works

  1. OpenClaw sends requests to https://api.naxxen.ai/nxn-sk-YOUR_KEY/v1/chat/completions (or /v1/responses, /v1/messages, etc.)
  2. naxxen strips the key from the path, compresses the prompt, and forwards to the real provider
  3. The response streams back through naxxen to OpenClaw

Both OpenAI endpoints (/v1/chat/completions and /v1/responses) are fully supported.

Configuration

Edit your OpenClaw config file (usually ~/.openclaw/openclaw.json). Add providers under models.providers — each provider is a keyed object (not an array entry).

The URL path prefix method embeds your naxxen key in the base URL. OpenClaw sends the provider API key in headers as usual.

Important notes

  • Provider entries go under models.providers as named keys (e.g. "naxxen-openai": { ... })
  • Each model needs id, name, contextWindow, and maxTokens
  • Set cost to 0 for all fields — naxxen tracks usage in its own dashboard
  • GPT-5.4 models need "compat": { "maxTokensField": "max_completion_tokens" }
  • The api field must match the provider type exactly

Full examples

Replace nxn-sk-YOUR_NAXXEN_KEY with your actual naxxen API key. Replace the provider API keys with your own.

OpenAI via naxxen

"naxxen-openai": {
  "baseUrl": "https://api.naxxen.ai/nxn-sk-YOUR_NAXXEN_KEY",
  "apiKey": "sk-proj-YOUR_OPENAI_KEY",
  "api": "openai-completions",
  "models": [
    {
      "id": "gpt-4o",
      "name": "GPT-4o (naxxen)",
      "api": "openai-completions",
      "reasoning": false,
      "input": ["text", "image"],
      "cost": { "input": 0, "output": 0, "cacheRead": 0, "cacheWrite": 0 },
      "contextWindow": 128000,
      "maxTokens": 16384
    },
    {
      "id": "gpt-4o-mini",
      "name": "GPT-4o Mini (naxxen)",
      "api": "openai-completions",
      "reasoning": false,
      "input": ["text", "image"],
      "cost": { "input": 0, "output": 0, "cacheRead": 0, "cacheWrite": 0 },
      "contextWindow": 128000,
      "maxTokens": 16384
    },
    {
      "id": "gpt-5.4",
      "name": "GPT-5.4 (naxxen)",
      "api": "openai-completions",
      "reasoning": false,
      "input": ["text", "image"],
      "cost": { "input": 0, "output": 0, "cacheRead": 0, "cacheWrite": 0 },
      "contextWindow": 1048576,
      "maxTokens": 32768,
      "compat": { "maxTokensField": "max_completion_tokens" }
    },
    {
      "id": "gpt-5.4-mini",
      "name": "GPT-5.4 Mini (naxxen)",
      "api": "openai-completions",
      "reasoning": false,
      "input": ["text", "image"],
      "cost": { "input": 0, "output": 0, "cacheRead": 0, "cacheWrite": 0 },
      "contextWindow": 1048576,
      "maxTokens": 32768,
      "compat": { "maxTokensField": "max_completion_tokens" }
    }
  ]
}

Anthropic via naxxen

"naxxen-anthropic": {
  "baseUrl": "https://api.naxxen.ai/nxn-sk-YOUR_NAXXEN_KEY",
  "apiKey": "sk-ant-YOUR_ANTHROPIC_KEY",
  "api": "anthropic-messages",
  "models": [
    {
      "id": "claude-opus-4-6",
      "name": "Claude Opus 4.6 (naxxen)",
      "api": "anthropic-messages",
      "reasoning": false,
      "input": ["text", "image"],
      "cost": { "input": 0, "output": 0, "cacheRead": 0, "cacheWrite": 0 },
      "contextWindow": 1000000,
      "maxTokens": 32768
    },
    {
      "id": "claude-sonnet-4-6",
      "name": "Claude Sonnet 4.6 (naxxen)",
      "api": "anthropic-messages",
      "reasoning": false,
      "input": ["text", "image"],
      "cost": { "input": 0, "output": 0, "cacheRead": 0, "cacheWrite": 0 },
      "contextWindow": 200000,
      "maxTokens": 8192
    },
    {
      "id": "claude-haiku-4-5-20251001",
      "name": "Claude Haiku 4.5 (naxxen)",
      "api": "anthropic-messages",
      "reasoning": false,
      "input": ["text", "image"],
      "cost": { "input": 0, "output": 0, "cacheRead": 0, "cacheWrite": 0 },
      "contextWindow": 200000,
      "maxTokens": 8192
    }
  ]
}

Google via naxxen

"naxxen-google": {
  "baseUrl": "https://api.naxxen.ai/nxn-sk-YOUR_NAXXEN_KEY",
  "apiKey": "YOUR_GOOGLE_API_KEY",
  "api": "google-generative-ai",
  "models": [
    {
      "id": "gemini-2.5-pro",
      "name": "Gemini 2.5 Pro (naxxen)",
      "api": "google-generative-ai",
      "reasoning": false,
      "input": ["text"],
      "cost": { "input": 0, "output": 0, "cacheRead": 0, "cacheWrite": 0 },
      "contextWindow": 1048576,
      "maxTokens": 8192
    },
    {
      "id": "gemini-2.5-flash",
      "name": "Gemini 2.5 Flash (naxxen)",
      "api": "google-generative-ai",
      "reasoning": false,
      "input": ["text"],
      "cost": { "input": 0, "output": 0, "cacheRead": 0, "cacheWrite": 0 },
      "contextWindow": 1048576,
      "maxTokens": 8192
    },
    {
      "id": "gemini-2.5-flash-lite",
      "name": "Gemini 2.5 Flash Lite (naxxen)",
      "api": "google-generative-ai",
      "reasoning": false,
      "input": ["text"],
      "cost": { "input": 0, "output": 0, "cacheRead": 0, "cacheWrite": 0 },
      "contextWindow": 1048576,
      "maxTokens": 8192
    }
  ]
}

Where to put it

These provider blocks go inside models.providers in your ~/.openclaw/openclaw.json:

{
  "models": {
    "providers": {
      "naxxen-openai": { ... },
      "naxxen-anthropic": { ... },
      "naxxen-google": { ... }
    }
  }
}

You can add naxxen providers alongside your existing direct providers. They don't replace anything — OpenClaw lets you pick which provider/model to use per request.

Allow the models

After adding providers, you must allow the naxxen models in your agent config. Under agents.defaults.models, add each model you want to use:

{
  "agents": {
    "defaults": {
      "models": {
        "naxxen-openai/gpt-5.4": {},
        "naxxen-openai/gpt-4o": {},
        "naxxen-anthropic/claude-sonnet-4-6": {},
        "naxxen-google/gemini-2.5-flash": {},

      }
    }
  }
}

Without this step, OpenClaw will reject the model with "Model is not allowed".

Using a naxxen model

Once configured and allowed, reference naxxen models with the provider prefix:

naxxen-openai/gpt-5.4
naxxen-anthropic/claude-sonnet-4-6
naxxen-google/gemini-2.5-flash

Or set one as your default model:

{
  "agents": {
    "defaults": {
      "model": {
        "primary": "naxxen-openai/gpt-5.4"
      }
    }
  }
}

Verify

After configuring, restart OpenClaw (openclaw gateway restart) and make a request using one of your naxxen providers. Then check your dashboard — you should see the request with compression stats.

Using OpenAI Responses API instead of Chat Completions?

If you're using OpenAI with the newer Responses API (/v1/responses), use "api": "openai-responses" instead of "openai-completions":

"naxxen-openai": {
  "baseUrl": "https://api.naxxen.ai/nxn-sk-YOUR_NAXXEN_KEY",
  "apiKey": "sk-proj-YOUR_OPENAI_KEY",
  "api": "openai-responses",
  "models": [
    {
      "id": "gpt-5.4",
      "name": "GPT-5.4 (naxxen)",
      "api": "openai-responses",
      "reasoning": false,
      "input": ["text", "image"],
      "cost": { "input": 0, "output": 0, "cacheRead": 0, "cacheWrite": 0 },
      "contextWindow": 1048576,
      "maxTokens": 32768,
      "compat": { "maxTokensField": "max_completion_tokens" }
    }
  ]
}

Most users should stick with "openai-completions" — it works with both API keys and the Chat Completions endpoint. Use "openai-responses" only if your setup specifically requires the Responses API.

Note: OpenAI Codex OAuth (ChatGPT subscription) uses a proprietary endpoint (chatgpt.com/backend-api) that cannot be proxied through naxxen. Codex OAuth is only supported as a direct connection via OpenClaw's built-in openai-codex provider.

Common mistakes

MistakeFix
Provider as array entryProviders are keyed objects under models.providers, not array items
Missing contextWindowEvery model needs contextWindow (in tokens)
Missing maxTokensEvery model needs maxTokens (max output tokens)
Wrong api valueMust be exactly openai-completions, anthropic-messages, or google-generative-ai
GPT-5.4 without compatGPT-5.4/5.4-mini need "compat": { "maxTokensField": "max_completion_tokens" }
Using name at provider levelProvider name is the key (e.g. "naxxen-openai"), not a field inside the object

Set up with an AI agent

Copy this prompt and give it to your AI coding agent:

Read https://docs.naxxen.ai/docs/integrations/openclaw and configure naxxen
as a proxy for my OpenAI, Anthropic, and Google API calls in OpenClaw.
My naxxen key is nxn-sk-YOUR_KEY.
Add the naxxen-openai, naxxen-anthropic, and naxxen-google providers
to my ~/.openclaw/openclaw.json config under models.providers.
Use the exact JSON examples from the docs — do not simplify or omit fields.

Compression

How naxxen compression benefits you and how to tune it.

Claude Code

Route Claude Code through naxxen for prompt compression.

On this page

How it worksConfigurationImportant notesFull examplesOpenAI via naxxenAnthropic via naxxenGoogle via naxxenWhere to put itAllow the modelsUsing a naxxen modelVerifyUsing OpenAI Responses API instead of Chat Completions?Common mistakesSet up with an AI agent