AI代理策略执行层

Moat

Intent Solutions is a Claude Partner Network Select Services Partner.

Policy-enforced execution and trust layer for AI agents. MCP-first.

Moat is not an API marketplace. It is the verified execution layer where agents call capabilities (method-level tool contracts), not connectors. Every call produces a receipt, an outcome event, and a policy decision record.

Interfaces: MCP (primary) + REST/OpenAPI (fallback) License: Elastic License 2.0

Core Guarantees

• Receipt on every call - append-only audit trail, idempotency, replay-ready
• Policy on every call - scopes, budgets, domain allowlists, default-deny
• Score on every call - success rate, latency percentiles, verified badges
• No open proxy - capabilities declare allowed outbound domains; everything else is blocked
• No raw secrets - credentials live in a vault; agents never see them

Architecture

Quick Start

git clone https://github.com/jeremylongshore/moat.git && cd moat
make install && bash scripts/dev.sh   # start all services
make demo                              # register → execute → receipt → stats

Docs

All documentation is in 000-docs/ using the doc-filing system. See the Standards Catalog.

Contributing

See CONTRIBUTING.md. Questions: Discussions. Security: SECURITY.md.