# ClawdRules

> AI agent governance platform. 482 configurable rules, behavioral anomaly detection, earned trust scores, soul.md personality profiles, and real-time spending guardrails. Built for the agentic commerce era.

## Docs

- [API Documentation](https://clawdrules.com/docs): Full REST API reference with code examples
- [OpenAPI Spec](https://clawdrules.com/openapi.yaml): OpenAPI 3.1 specification for all endpoints
- [Integrations](https://clawdrules.com/integrations): MCP server, platform support, auto-discovery
- [Plugin Manifest](https://clawdrules.com/.well-known/ai-plugin.json): Multi-platform agent plugin manifest

## API Endpoints

- [POST /connect](https://clawdrules.com/docs#connect): Register agent and load operating context (rules, soul, merchants, trust level)
- [POST /check](https://clawdrules.com/docs#check): Pre-flight transaction check — allow, deny, or escalate with anomaly scoring
- [GET /rules](https://clawdrules.com/docs#rules): Retrieve configured governance rules
- [GET /soul](https://clawdrules.com/docs#soul): Retrieve agent personality profile (soul.md)
- [GET /stores](https://clawdrules.com/docs#stores): Retrieve trusted and blocked merchant lists
- [GET /trust](https://clawdrules.com/docs#trust): Get current trust score, level, multiplier, and baseline
- [POST /checkout](https://clawdrules.com/docs#checkout): Create, complete, cancel checkout sessions — ACP lifecycle
- [POST /checkout/receipt](https://clawdrules.com/docs#receipt): Report purchase receipt after external payment

## Capabilities

- Rule engine: 482 rules across 5 security domains (spending limits, merchant controls, velocity checks, budget categories, behavioral boundaries)
- Anomaly detection: Per-agent behavioral baselines computed from 30-day transaction history. Composite 0-100 anomaly score from 5 weighted flags (amount z-score, velocity, time-of-day, merchant novelty, category novelty). Informational only — never auto-denies.
- Trust scores: Earned autonomy model. Agents start at 50 and gain or lose trust from behavior. Trust level determines spending multiplier. Levels: untrusted (0-19), low (20-39), medium (40-59), high (60-79), very_high (80-100).
- Soul.md: Personality profiles with 80+ traits, 70+ behavioral directives, 6 archetypes (Household Helper, Shopping Pro, Family Guardian, Budget Warrior, Privacy Shield, Custom). Defines agent voice, values, and origin story.
- Commerce governance: ACP (Agentic Commerce Protocol) and AP2 (Agent Payment Protocol) support. Full checkout lifecycle: create → evaluate → approve/deny/escalate → pay externally → report receipt.
- Kill switch: Instant freeze of all agent activity. Per-agent and global (per-user) kill switches.
- Notifications: Multi-channel alerts (Slack, Teams, webhook, email) for anomalies, spending digests, and governance events.

## Integration

- MCP Server: `claude mcp add clawdrules -- npx -y clawdrules-mcp` — 13 governance tools for Claude and MCP-compatible clients
- REST API: Bearer token auth (`Authorization: Bearer clr_live_...`), base URL `https://clawdrules.com/api/v1/agent/`
- Auto-discovery: `/.well-known/ai-plugin.json` with extensions for Anthropic, Google Gemini, xAI Grok
- Supported platforms: Claude/Anthropic (MCP native), Google Gemini (Gemini Actions), xAI Grok (Grok Actions), OpenAI/ChatGPT (REST API), any HTTP-capable agent

## Getting Started

- [Create Account](https://clawdrules.com/register): Free, passkey-only authentication (no passwords)
- [Sign In](https://clawdrules.com/login): WebAuthn passkey login