# Claw-Family AI Agent Integration

Swan Inference provides an OpenAI-compatible API that works with all Claw-family AI agent tools. This guide covers how to connect each tool to Swan Inference for decentralized AI inference.

## Prerequisites

1. A Swan Inference API key (`sk-swan-*`) — [sign up here](https://inference.swanchain.io/signup)
2. The Claw tool of your choice installed

## Quick Start

All Claw-family tools support custom OpenAI-compatible endpoints. The core configuration is the same across all tools:

| Setting  | Value                                                                                 |
| -------- | ------------------------------------------------------------------------------------- |
| Base URL | `https://inference.swanchain.io/v1`                                                   |
| API Key  | `sk-swan-YOUR-API-KEY`                                                                |
| Model    | Any model from [inference.swanchain.io/models](https://inference.swanchain.io/models) |

Popular models available on Swan Inference:

| Model                               | Category | Use Case                     |
| ----------------------------------- | -------- | ---------------------------- |
| `deepseek-r1-distill-llama-70b`     | LLM      | Reasoning, code generation   |
| `Qwen/Qwen2.5-7B-Instruct`          | LLM      | General chat, fast responses |
| `meta-llama/Llama-3.3-70B-Instruct` | LLM      | General purpose              |

***

## Tool-Specific Setup

### OpenClaw (TypeScript, 250K+ stars)

The most popular AI assistant in the Claw family. Supports 25+ messaging platforms.

{% tabs %}
{% tab title="Config File" %}
Edit your OpenClaw configuration to add Swan Inference as a provider:

```json
{
  "models": {
    "providers": {
      "swan": {
        "api": "openai-completions",
        "baseUrl": "https://inference.swanchain.io/v1",
        "apiKey": "sk-swan-YOUR-API-KEY",
        "models": [
          "deepseek-r1-distill-llama-70b",
          "Qwen/Qwen2.5-7B-Instruct"
        ]
      }
    },
    "default": "swan:deepseek-r1-distill-llama-70b"
  }
}
```

{% endtab %}

{% tab title="Environment Variables" %}

```bash
export OPENAI_API_BASE=https://inference.swanchain.io/v1
export OPENAI_API_KEY=sk-swan-YOUR-API-KEY
openclaw
```

{% endtab %}
{% endtabs %}

***

### ZeroClaw (Rust, 28K+ stars)

Ultra-lightweight (3.4MB binary, <10ms startup). Best for quick testing.

{% tabs %}
{% tab title="config.toml" %}

```toml
[provider]
name = "openai-compatible"
base_url = "https://inference.swanchain.io/v1"
api_key = "sk-swan-YOUR-API-KEY"
model = "deepseek-r1-distill-llama-70b"
```

{% endtab %}

{% tab title="CLI" %}

```bash
zeroclaw --provider "custom:https://inference.swanchain.io/v1" \
         --api-key "sk-swan-YOUR-API-KEY" \
         --model "deepseek-r1-distill-llama-70b"
```

{% endtab %}
{% endtabs %}

***

### PicoClaw (Go, 26K+ stars)

Lightweight Go binary (<8MB). Designed for edge and ARM devices.

```yaml
# config.yaml
provider:
  type: openai-compatible
  base_url: https://inference.swanchain.io/v1
  api_key: sk-swan-YOUR-API-KEY
  model: deepseek-r1-distill-llama-70b
```

***

### Nanobot (Python, 32K+ stars)

Python-native, installable via pip. Best for Python developers.

{% tabs %}
{% tab title="Install" %}

```bash
pip install nanobot-ai
```

{% endtab %}

{% tab title="Config" %}

```json
{
  "provider": "custom",
  "apiBase": "https://inference.swanchain.io/v1",
  "apiKey": "sk-swan-YOUR-API-KEY",
  "model": "deepseek-r1-distill-llama-70b"
}
```

{% endtab %}

{% tab title="Python API" %}

```python
from nanobot import Agent

agent = Agent(
    provider="custom",
    api_base="https://inference.swanchain.io/v1",
    api_key="sk-swan-YOUR-API-KEY",
    model="deepseek-r1-distill-llama-70b",
)

response = agent.chat("Explain quantum computing in simple terms")
print(response)
```

{% endtab %}
{% endtabs %}

***

### NanoClaw (TypeScript, 25K+ stars)

Container-per-session isolation. Good for multi-tenant deployments.

```json
{
  "llm": {
    "provider": "openai-compatible",
    "baseUrl": "https://inference.swanchain.io/v1",
    "apiKey": "sk-swan-YOUR-API-KEY",
    "model": "deepseek-r1-distill-llama-70b"
  }
}
```

***

### IronClaw (Rust, 10K+ stars)

Security-focused with TEE and encrypted vault. Built by NEAR AI.

```bash
export LLM_BACKEND=openai_compatible
export LLM_BASE_URL=https://inference.swanchain.io/v1
export API_KEY=sk-swan-YOUR-API-KEY
export MODEL=deepseek-r1-distill-llama-70b

ironclaw
```

***

### NemoClaw (TypeScript + Python, 15K+ stars)

NVIDIA's reference stack with kernel sandbox and privacy router.

Configure the inference provider in your NemoClaw deployment:

```yaml
inference:
  provider: openai-compatible
  base_url: https://inference.swanchain.io/v1
  api_key: sk-swan-YOUR-API-KEY
  model: deepseek-r1-distill-llama-70b
```

{% hint style="info" %}
NemoClaw is in early preview (alpha). Configuration may change.
{% endhint %}

***

### Moltworker (TypeScript, Cloudflare Workers)

Runs on Cloudflare's edge network (330+ cities).

```json
{
  "models": {
    "providers": {
      "swan": {
        "type": "openai-compatible",
        "baseUrl": "https://inference.swanchain.io/v1",
        "apiKey": "sk-swan-YOUR-API-KEY"
      }
    }
  }
}
```

Or set via Cloudflare AI Gateway:

```bash
AI_GATEWAY_BASE_URL=https://inference.swanchain.io/v1
```

***

### NullClaw (Zig, 6.7K+ stars)

Ultra-minimal (678KB binary, 1MB RAM). For IoT and embedded devices.

```bash
nullclaw --provider "custom:https://inference.swanchain.io/v1" \
         --api-key "sk-swan-YOUR-API-KEY" \
         --model "Qwen/Qwen2.5-7B-Instruct"
```

Or in config:

```toml
[provider]
url = "custom:https://inference.swanchain.io/v1"
api_key = "sk-swan-YOUR-API-KEY"
model = "Qwen/Qwen2.5-7B-Instruct"
```

{% hint style="info" %}
For resource-constrained devices, use smaller models like `Qwen/Qwen2.5-7B-Instruct` for faster responses.
{% endhint %}

***

## Choosing the Right Tool

| Scenario            | Recommended Tool | Why                                |
| ------------------- | ---------------- | ---------------------------------- |
| Quick local testing | **ZeroClaw**     | 3.4MB, boots in <10ms              |
| Python project      | **Nanobot**      | pip install, native Python API     |
| Multi-platform bot  | **OpenClaw**     | 25+ messaging platforms            |
| Edge / ARM device   | **PicoClaw**     | 8MB Go binary, runs anywhere       |
| IoT / embedded      | **NullClaw**     | 678KB, runs on $5 hardware         |
| Security-critical   | **IronClaw**     | TEE, encrypted vault, WASM sandbox |
| Multi-tenant SaaS   | **NanoClaw**     | Container-per-session isolation    |
| Enterprise / NVIDIA | **NemoClaw**     | Kernel sandbox, policy enforcement |
| Global edge deploy  | **Moltworker**   | Cloudflare Workers, 330+ cities    |

## Playground (No API Key)

All tools can also use Swan Inference's public playground for testing without an API key:

```
Base URL: https://inference.swanchain.io/v1/playground
```

The playground is rate-limited (5 requests/hour per IP) with restricted models, but requires no signup.

## Troubleshooting

### Connection refused / timeout

* Verify the base URL includes `/v1`: `https://inference.swanchain.io/v1`
* Check your API key starts with `sk-swan-`

### Model not found

* List available models: `curl https://inference.swanchain.io/v1/models -H "Authorization: Bearer sk-swan-YOUR-KEY"`
* Model IDs are case-sensitive (e.g., `Qwen/Qwen2.5-7B-Instruct`, not `qwen2.5-7b`)

### Rate limited (429)

* Default rate limit is 200 requests/min for LLM models
* Check `X-RateLimit-Remaining` header in responses
* Consider upgrading to Pro subscription ($6/month) for higher limits

### Streaming not working

* Ensure your tool is configured for streaming (`stream: true`)
* Swan Inference supports SSE streaming on `/v1/chat/completions`

## Learn More

* [Swan Inference API Reference](/bulders/app-developer/swan-inference-api.md)
* [Available Models](https://inference.swanchain.io/models)
* [Sign Up](https://inference.swanchain.io/signup)
* [Subscription Plans](/bulders/app-developer/swan-inference-api.md#subscription-plan)


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://docs.swanchain.io/bulders/app-developer/claw-tools-integration.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.