Assistant (default)
Chat-based assistant with smart escalation
routing_mode: "assistant"Get Started
Pick your mode. Set your budget. Get your config. Run.
How will you use Toolklaw?
Reference
Pre-built configs for common use cases.
Chat-based assistant with smart escalation
routing_mode: "assistant"Robot control, ultra-cheap, no premium
routing_mode: "builder"Robot with safety gates and structured output
routing_mode: "embodied"Help
How to plug this config into your gateway.
pip install toolklaw-clientproviders:
toolklaw:
type: "llm"
config: |
provider: toolklaw
model: tk_claw
routing_mode: assistant
# ... paste your config hereexport TOOLKLAW_KEY="tkc_live_..."python3 -m openclaw.gateway --config gateway.yamlcurl -X POST http://localhost:8000/message \
-d '{"text": "Hello"}' \
-H "Content-Type: application/json"Troubleshooting
How to fix things if they don't work.
Check that your TOOLKLAW_KEY is set and valid. Get a new one from your dashboard.
Make sure your config has routing_mode: "assistant" (not a typo). Test with a simple prompt first.
Increase your max_per_turn_cents in the config. Some prompts need more. Check your usage in the dashboard.
Add your GPT key to dashboard Settings → Providers. Check that byo_enabled: true in your config.
tk_claw usually responds in <500ms. If you're hitting timeouts, check your network. File an issue in Discord.
Join our Discord. Operators help operators. Or email support@toolklaw.com.
Your config is ready. Get an API key and start building.