ao-mcp-tollbooth

🛡️ AO MCP Tollbooth A Model Context Protocol (MCP) server providing automated risk triage, liquidity checks, and security guardrails for AI agents operating on the AO / Arweave network.

🌐 Community & Ecosystem Integration ao-mcp-tollbooth is registered as a specialized Community Server for autonomous agents across the Web3 and AO ecosystems.

Registry Entry: * ao-mcp-tollbooth - Automated risk triage and security guardrails for AI agents on AO/Arweave.

Category: Security, Risk Triage & Infrastructure

Target Network: AO / Arweave Ecosystem

🚀 Overview ao-mcp-tollbooth acts as an automated safety gate for autonomous agents. Before executing financial interactions or interacting with unknown processes on AO, agents pass process metadata to Tollbooth to receive a real-time risk assessment and actionable trigger responses.

Key Features GraphQL Process Triage: Fetches process history and transaction metadata directly via Arweave gateways.

Financial Risk Heuristics: Computes a normalized financial_risk_score (0–100) based on verified credentials, transaction volume, and interaction history.

Agent Decision Triggers: Emits standardized actions (ALLOW, FLAG_FOR_REVIEW, BLOCK) for seamless integration with LLM decision loops.

Zero Overhead: Pure local computation for triage execution with minimal network latency.

🛠️ Quickstart Prerequisites Python 3.10+

Git

Installation Bash git clone https://github.com/Ryddegutt/ao-mcp-tollbooth.git cd ao-mcp-tollbooth pip install -e . 💡 Usage Start the MCP server locally:

Bash python server.py Response Schema Example JSON { "process_id": "0x123...abc", "alpha_score": 85, "financial_risk_score": 20, "is_verified": true, "agent_action": "ALLOW" } ⚖️ Disclaimer This software provides heuristic risk scoring based on publicly available ledger metadata. It does not constitute financial or legal advice, nor does it guarantee the absolute safety of target processes.

📜 License MIT License © 2026 Ryddegutt