🛡️ SentinelMesh

SentinelMesh is a real-time security, policy, and observability mesh for AI agents communicating via the Model Context Protocol (MCP).

🚀 Features

• 🔎 Real-time monitoring of MCP agent-to-agent messages
• 🔐 Prompt injection & sensitive context detection (YAML-defined rules)
• ⚠️ Risk scoring with visual color tags
• 📊 Streamlit dashboard with filters, charts, and risk summaries
• 🧪 Log replay tool for forensic analysis
• 🛑 Manual + auto-quarantine via blocklist
• 🧱 Modular CLI structure (installed with setup.py)

📦 Installation

From your project root:

pip install -e .

🖥️ Dashboard

Launch the real-time Streamlit dashboard:

streamlit run sentinelmesh/dashboard/app.py

Features:

• Realtime log table
• Risk scoring with heatmap
• Agent volume and risk timeline
• Block/unblock agents live from the sidebar

🧰 CLI Tools

MCP Monitor:

sentinelmesh-cli

Monitors the log file and shows alerts in real time.

Log Replay:

sentinelmesh-replay --agent agent.hrbot.local --limit 10

Replay historical logs with filters.

⚙️ Config Files

• sentinelmesh/rules/rules.yaml — security rules (phrases, context, regex)
• sentinelmesh/rules/blocklist.yaml — blocked agents (manually set)
• sentinelmesh/logs/mcp_traffic.log — live traffic log

📝 License

MIT © Rishit Goel