@agentlaw/gateway

The agentlaw CLI and MCP policy gateway — the enforcement point of AgentLaw. Wraps any MCP server behind a signed policy manifest: every tool call is checked by the deterministic engine in @agentlaw/core and allowed, paused for human approval, or denied — fail-closed, hash-chain audited.

Quickstart

npx @agentlaw/gateway wizard .

Scaffolds a signed, ready-to-run policy.yaml (template + thresholds + keypair) and prints the .mcp.json snippet to wrap your MCP servers:

{
  "mcpServers": {
    "safe-fs": {
      "command": "npx",
      "args": ["-y", "@agentlaw/gateway", "--policy", "policy.yaml", "--trust", "policy.pub",
               "--", "npx", "-y", "@modelcontextprotocol/server-filesystem", "/sandbox"]
    }
  }
}

CLI

| Command | What it does | | --- | --- | | agentlaw wizard . | Guided setup: template, thresholds, signed policy + keypair. | | agentlaw sign | Sign a manifest with your Ed25519 private key. | | agentlaw verify | Verify a manifest's signature against a trusted public key. | | agentlaw explain | Dry-run: the verdict a tool call would receive, naming the rule. | | agentlaw audit verify | Check the integrity of the hash-chained audit log. |

Human-in-the-loop approve rules prompt on the enforcement host's own terminal — no external service; no terminal means deny, never silent allow.

Spec and threat model: RFC-0001.

Apache-2.0.