How to Integrate Together AI Models in Openclaw

Table of contents: [Show]

Together AI Openclaw Integration

Together AI provides access to high-performance open-source models like Kimi K2.5 through a fast, scalable inference API. This guide walks you through integrating Together AI models into your Openclaw setup.

Quick Start

Before you begin, grab your TOGETHER_API_KEY from the Together AI dashboard. You'll need this to authenticate requests.

Interactive Setup

Run the Openclaw onboarding wizard and select Together AI when prompted:

openclaw onboard --auth-choice together-api-key

Paste your API key when prompted. The wizard will configure your ~/.openclaw/openclaw.json automatically.

Non-Interactive Setup

For CI/CD or automated deployments, use the non-interactive mode:

openclaw onboard --non-interactive --mode local --auth-choice together-api-key --together-api-key "$TOGETHER_API_KEY"

Configure Your Primary Model

Edit your ~/.openclaw/openclaw.json to set Kimi K2.5 as your default model:

{
  "agents": {
    "defaults": {
      "model": { "primary": "together/moonshotai/Kimi-K2.5" }
    }
  }
}

If you're exploring other AI providers, check out our guides on OpenRouter Models and LiteLLM Models for more options.

Environment Setup for Daemon Mode

If you run Openclaw Gateway as a background service via systemd or launchd, ensure the API key is accessible:

echo "TOGETHER_API_KEY=your_key_here" >> ~/.openclaw/.env

Alternatively, configure env.shellEnv in your Openclaw config to export the variable.

Test Your Setup

Verify the integration works by sending a test prompt:

openclaw ask "Explain quantum computing in one sentence"

You should receive a response from Kimi K2.5 via Together AI's infrastructure.

Troubleshooting

401 Unauthorized: Double-check your API key is valid and not expired. Regenerate at the Together AI dashboard if needed.
Model not found: Ensure you're using the correct model ID prefix together/ followed by the full path like moonshotai/Kimi-K2.5.
Rate limits: Together AI enforces rate limits based on your plan. Check your usage dashboard if requests fail with 429 errors.

Best Practices

Store API keys in ~/.openclaw/.env rather than hardcoding in config files
Use --non-interactive mode for containerized deployments
Monitor token usage through the Together AI dashboard to optimize costs

LLM Task Integration

Ready to explore more AI integrations? Browse our Venice AI setup guide for privacy-focused inference options.