# AgenticRail Gate - sequence enforcement and verifiable audit receipts for AI agent compliance

> AgenticRail Gate - sequence enforcement and verifiable audit receipts for AI agent compliance is a freemium MCP server for security. Step-order enforcement for AI agents: ALLOW or DENY each step, with signed receipts. It runs as a hosted remote server, so there is nothing to install locally.

- **Kind:** MCP server
- **Pricing:** Freemium
- **Starting price:** $39 a month
- **Platforms:** API
- **Website:** https://agenticrail.nz/docs/
- **Repository:** https://github.com/msmd-rua/agenticrail-mcp
- **MCP Registry name:** nz.agenticrail/gate
- **Categories:** security, AI models and LLM tooling
- **Listing updated:** 2026-11-16

## About

The gate works by making an agent ask permission before every step. The agent posts its current step with a unique nonce and timestamp, and the gate checks the step order, rejects replays, confirms the action type is allowed for that step and confirms the sequence has not been sealed. Pass gives ALLOW and a receipt; failure gives DENY with a reason code, and an agent is meant to proceed only on ALLOW. The README frames this as catching skipped steps and procedural hallucination, where an agent claims a step succeeded that never ran.

Every decision becomes an Ed25519-signed, chained receipt written as a tamper-evident record, and a second tool fetches a verification report to confirm the chain is intact. Denials return their own remedy, such as the allowed action types or the next expected step, and a refused call locks nothing, so the sequence is not spoiled. Action types are limited to eight values, and the README notes invented verbs like READ or EXECUTE are the most common first refusal.

A public demo key lets you test without signing up, in a shared and rate-limited lane. A private developer key costs US$39 a month, and deployments are arranged by email.

## Features

- **Pre-step gating:** Each step is allowed or denied before it runs.
- **Signed receipts:** Ed25519-signed, chained records for every decision.
- **Replay protection:** Nonces prevent a step being replayed.
- **Self-explaining denials:** A refusal returns allowed action types or the next expected step.
- **Custom step order:** Send your own step_order and the gate enforces it.
- **Verification report:** Check that a sequence's receipt chain is intact.

## Good for

- Stop an agent from skipping an identity check before a transaction
- Keep an audit trail of what an agent was allowed to do
- Detect agents that report success on a step that never ran
- Produce compliance evidence for an agent workflow

## Install

### Claude Code (Terminal)

```sh
claude mcp add --transport http gate https://mcp.agenticrail.nz/
```

### Claude Desktop (Settings → Connectors → Add custom connector)

```sh
https://mcp.agenticrail.nz/
```

### Cursor (~/.cursor/mcp.json)

```json
{
  "mcpServers": {
    "gate": {
      "url": "https://mcp.agenticrail.nz/"
    }
  }
}
```

### VS Code (.vscode/mcp.json)

```json
{
  "servers": {
    "gate": {
      "type": "http",
      "url": "https://mcp.agenticrail.nz/"
    }
  }
}
```

### Windsurf (~/.codeium/windsurf/mcp_config.json)

```json
{
  "mcpServers": {
    "gate": {
      "serverUrl": "https://mcp.agenticrail.nz/"
    }
  }
}
```

### Other clients (your client's MCP settings (most use this shape))

```json
{
  "mcpServers": {
    "gate": {
      "type": "http",
      "url": "https://mcp.agenticrail.nz/"
    }
  }
}
```

## Tools

- `evaluate_step`: ALLOW or DENY a single agent step before it runs and seal a signed receipt.
- `verify_receipt`: Fetch a sequence's verification report and confirm the receipt chain is intact.

## FAQ

### Can I try it without signing up?

Yes, a public demo key works in a rate-limited shared lane.

### What does a private key cost?

US$39 a month, bought without speaking to anyone.

### What happens on a denial?

The gate returns a reason code and the remedy, and a refused call does not lock the sequence.

### How many tools does the server expose?

Two: evaluate_step and verify_receipt.

### Is AgenticRail Gate - sequence enforcement and verifiable audit receipts for AI agent compliance free?

AgenticRail Gate - sequence enforcement and verifiable audit receipts for AI agent compliance has a free plan, with paid plans for more usage or features. Paid plans start at $39 a month.

### How do I add AgenticRail Gate - sequence enforcement and verifiable audit receipts for AI agent compliance to Claude Code?

Run this command in your terminal: claude mcp add --transport http gate https://mcp.agenticrail.nz/

### What are the best alternatives to AgenticRail Gate - sequence enforcement and verifiable audit receipts for AI agent compliance?

The closest alternatives to AgenticRail Gate - sequence enforcement and verifiable audit receipts for AI agent compliance are REA, Claude Code Ultimate Guide, Agent Roast, Midplane and Leash.

## Alternatives

- [REA](https://ai-powered.com/mcp/rea): REA is an open-source MCP server for developer tools.
- [Claude Code Ultimate Guide](https://ai-powered.com/mcp/claude-code-ultimate-guide): Claude Code Ultimate Guide is an open-source MCP server for ai models and llm tooling.
- [Agent Roast](https://ai-powered.com/mcp/agent-roast): Agent Roast is an open-source MCP server for developer tools.
- [Midplane](https://ai-powered.com/mcp/midplane): Midplane is an open-source MCP server for databases.
- [Leash](https://ai-powered.com/mcp/leash): Leash is a free MCP server for security.
- [SecureCloud](https://ai-powered.com/mcp/securecloud): SecureCloud is a free MCP server for security.
- [Tinplate Edge AI Safety, Security & EU Compliance](https://ai-powered.com/mcp/tinplate-edge-ai-safety-security-and-eu-compliance): Tinplate Edge AI Safety, Security & EU Compliance is a freemium MCP server for security.
- [Eikino](https://ai-powered.com/mcp/eikino): Eikino is a freemium MCP server for security.

---
Source: https://ai-powered.com/mcp/agenticrail-gate-sequence-enforcement-and-verifiable-audit-r
