OpenClaw

Manual Configuration

If you prefer to configure OpenClaw manually, add ANT to your ~/.openclaw/openclaw.json:

{
  "providers": {
    "antbase": {
      "baseUrl": "https://antbase.ai",
      "apiKey": "${ANT_API_KEY}",
      "api": "anthropic-messages",
      "models": [
        { "id": "auto", "name": "ANT Auto", "reasoning": true, "input": ["text"], "contextWindow": 200000, "maxTokens": 8192 },
        { "id": "best", "name": "ANT Best", "reasoning": true, "input": ["text"], "contextWindow": 200000, "maxTokens": 16384 },
        { "id": "fast", "name": "ANT Fast", "reasoning": false, "input": ["text"], "contextWindow": 128000, "maxTokens": 4096 },
        { "id": "free", "name": "ANT Free", "reasoning": false, "input": ["text"], "contextWindow": 131072, "maxTokens": 8192 },
        { "id": "reasoning", "name": "ANT Reasoning", "reasoning": true, "input": ["text"], "contextWindow": 200000, "maxTokens": 16384 },
        { "id": "coder", "name": "ANT Coder", "reasoning": true, "input": ["text"], "contextWindow": 200000, "maxTokens": 16384 },
        { "id": "creative", "name": "ANT Creative", "reasoning": false, "input": ["text"], "contextWindow": 200000, "maxTokens": 16384 },
        { "id": "vision", "name": "ANT Vision", "reasoning": false, "input": ["text", "image"], "contextWindow": 128000, "maxTokens": 8192 }
      ]
    }
  }
}

Get your API key from the dashboard and set it in your environment:

export ANT_API_KEY="ant_your-api-key-here"

How Routing Works

When OpenClaw sends a request through ANT, it passes through a multi-stage pipeline before reaching a model:

┌──────────┐   ┌─────────────┐   ┌──────────────┐   ┌──────────┐   ┌──────────┐
│ OpenClaw │──▶│ Classify LLM│──▶│ Optimize LLM │──▶│ Route AI │──▶│   LLM    │
└──────────┘   └─────────────┘   └──────────────┘   └──────────┘   └──────────┘
                Task type &        Compress           Score &        2,500+ models
                complexity         (optional)         select         30+ providers

1. Classify — ANT's custom classifier tags the request by task type (coding, reasoning, creative, etc.), complexity (simple/medium/complex), and required capabilities (vision, tools, reasoning)
2. Optimize — Optional prompt compression removes redundant tokens, saving 20-40% on token costs
3. Route — Scores all available models using preference weights (speed, quality, cost) combined with task classification
4. Execute — Forwards to the highest-scoring model. If the provider fails, ANT re-scores and falls back to the next-best model from a different provider

The full pipeline adds ~100-200ms of overhead. For detailed scoring formula and configuration, see How It Works.

Virtual Models

Virtual models are routing presets. Instead of picking a specific model, you tell ANT what you care about and it selects the real model for each request. Use different virtual models for different use cases:

For the full list of virtual models, weight configurations, and usage examples, see Virtual Models.

Custom Routing with Nests

Nests let you create custom routing groups that restrict which models ANT can select from. Assign a nest to an API key, and all requests using that key route exclusively through the nest's model set — with your custom preference weights.

Create and manage nests in the dashboard under Models > Nests. See Nests documentation for full setup instructions.

Usage

Once configured, OpenClaw works exactly as before. ANT handles routing transparently:

# Your existing OpenClaw commands work as before
openclaw chat "How do I implement a binary search?"

# ANT will:
# 1. Classify the prompt (coding, medium complexity)
# 2. Select the optimal model (e.g., DeepSeek V3 for code)
# 3. Stream the response back with routing info in headers

# Switch models per-request:
openclaw -m antbase/free "explain what this regex does"
openclaw -m antbase/best "design the database schema for a multi-tenant SaaS"
openclaw -m antbase/reasoning "prove that sqrt(2) is irrational"

ANT returns routing metadata in response headers: X-Antbase-Model (actual model used), X-Antbase-Task-Type (detected task), X-Antbase-Complexity (simple/medium/complex).

Automatic Failover

If the selected provider returns an error or times out, ANT automatically fails over to the next-best scored model from a different provider. Circuit breakers track provider health in real-time and temporarily remove unhealthy providers from the scoring pool.

Troubleshooting