Recovery infrastructure for AI agents. When your tool calls fail, AgentRx tells you exactly what to do next.
Copy this folder to your OpenClaw skills directory: cp -r agentrx-openclaw ~/.openclaw/skills/agentrx
Or install via ClawHub once published: npx clawhub install agentrx
Add these to ~/.openclaw/.env: AGENTRX_API_KEY=your_api_key_here AGENTRX_BASE_URL=https://agentrx-production.up.railway.app
Beta API key (free tier, rate limited): AGENTRX_API_KEY=beta_openclaw_try_agentrx_2026
This shared key lets you test immediately. For a dedicated production key with higher limits, email chainassetslab@gmail.com
Classifies tool failures into 10 signatures and returns a direct plaintext instruction in openclaw_instruction telling the agent exactly what to do next. No interpretation required.
Failure signatures: AGENT_LOOP, AUTH_FAILURE, RATE_LIMIT_EXCEEDED, NETWORK_LATENCY, HALLUCINATED_PARAM, HALLUCINATED_VALUE, SCHEMA_MISMATCH, RESOURCE_MISSING, TOOL_DEPRECATED, UNKNOWN
After any tool failure: bash scripts/recover.sh "agent_id" "tool_name" "error message" 0
Before risky tool calls: bash scripts/preflight.sh "agent_id" "tool_name" '{"key":"val"}' '{}'
Live API: https://agentrx-production.up.railway.app/docs PyPI SDK: https://pypi.org/project/agentrx-sdk GitHub: https://github.com/chainassetslab/agentrx Contact: chainassetslab@gmail.com