Cloud Backends

Cloud backends send tool call context to a remote API for scoring. Faster model options, but requires an API key and sends data off-device.

OpenRouter

OpenRouter provides access to dozens of models through a single API key — Claude, GPT-4o, Gemini, Mistral, and more.

guardclaw config set SAFEGUARD_BACKEND openrouter
guardclaw config set OPENROUTER_API_KEY sk-or-...
guardclaw config set OPENROUTER_MODEL openai/gpt-4o-mini

Recommended models for cost/accuracy balance:

• openai/gpt-4o-mini — fast, cheap, accurate
• anthropic/claude-3-haiku — excellent reasoning
• google/gemini-flash-1.5 — very fast

---

Anthropic Claude

Direct Anthropic API. Highest accuracy for complex tool call analysis.

guardclaw config set SAFEGUARD_BACKEND anthropic
guardclaw config set ANTHROPIC_API_KEY sk-ant-...

GuardClaw uses claude-haiku-4-5-20251001 by default (fast, low cost). Upgrade to Sonnet for maximum accuracy:

guardclaw config set ANTHROPIC_MODEL claude-sonnet-4-6

---

MiniMax

MiniMax offers large-context models at competitive pricing.

guardclaw config set SAFEGUARD_BACKEND lmstudio
guardclaw config set LMSTUDIO_URL https://api.minimaxi.chat/v1
guardclaw config set LMSTUDIO_API_KEY your-minimax-key
guardclaw config set LMSTUDIO_MODEL MiniMax-M2.7

Available models:

Model	Context	Notes
MiniMax-M2.7	205K	Best quality
MiniMax-M2.5	205K	Balanced
MiniMax-Text-01	1M	Largest context

---

Any OpenAI-compatible API

GuardClaw's lmstudio backend accepts any OpenAI-compatible endpoint. This covers DeepSeek, Groq, Fireworks, Together AI, and others:

guardclaw config set SAFEGUARD_BACKEND lmstudio
guardclaw config set LMSTUDIO_URL https://api.groq.com/openai/v1
guardclaw config set LMSTUDIO_API_KEY gsk_...
guardclaw config set LMSTUDIO_MODEL llama-3.1-70b-versatile