Agent Action Gate MCP server
What it does
Agent Action Gate is an MCP server that gates and audits agent tool calls before execution, with a focus on blocking unattended destructive operations like shell execution and decommissioning. It enforces human-in-the-loop approval for high-risk actions and maintains an append-only audit ledger, designed for teams building AI agents who need to demonstrate control compliance.
- Deny shell execution and infrastructure decommissioning commands unless explicitly approved by a human with a prove token.
- Simulate write operations to detect unintended side effects before they reach production.
- Generate audit trail records of all gated decisions with hash-chained ledger entries for compliance audits and diligence reviews.
Tools this server exposes
| Tool | Does |
|---|---|
| gate_check | Evaluates an MCP tool call against Gate/Prove policy and returns allow, deny, or simulate mode with reasoning. |
| ledger_verify | Retrieves and verifies hash-chained action ledger entries for audit and compliance purposes. |
Before you enable it
An MCP server runs with whatever access you give it. Start with a development account rather than production, and check which tools the server exposes before you let an agent call them unattended.