Vision & team
Paart is built on one belief: you should see and control exactly what your AI agents do, on infrastructure you own.
Security for the agent era should be local, sovereign, and honest.
AI agents are becoming the largest and fastest-growing class of identities acting on our systems, and the only one that decides for itself. Copilots, IDE assistants, autonomous agents and MCP servers now make real tool calls against real data, on your behalf.
Our belief is simple: you should see and control exactly what those agents do, on infrastructure you own, without shipping your data to anyone. Trust should come from running locally, from an open foundation, and from honest claims, not from marketing.
So we build mcpids, the open-source MCP security gateway, in the open, and Paart as the product that turns that visibility into enforcement you can trust.
Runs on your machine. Your agent data does not leave it by default.
Built and run in the EU. No dependency on a foreign control plane.
Inspects MCP, OpenAI and Anthropic tool calls at the wire.
Every capability tagged stable, experimental or planned. No overclaiming.
Deep security and applied-ML experience, kept deliberately lean. We are growing.
Cengis Hasan
Ex-Huawei. PhD; 15+ years in ML and multi-agent systems. CEO of Cognifinity.
LinkedIn
