Cognis

The governed runtime for AI agents

2 followers

The governed runtime for AI agents

2 followers

Cognis is an open-source, local-first runtime for executing AI agent workflows with policy enforcement, human authorization, controlled MCP tool execution, and audit logging.

Free

Launch tags:Productivity•Developer Tools•Artificial Intelligence

Launch Team

Maker

📌

Hey everyone — I'm Aritra, the person building Cognis. I've been working on the project because I think there's a missing layer between AI agents and the systems they can control. MCP is making agent/tool interoperability much easier, but I wanted to explore what happens when you add an actual governed execution layer around it: policies, human authorization, security boundaries, and an audit trail. Cognis is my attempt at building that layer as an open-source project. It's still early, so I'm particularly interested in feedback from people building agents, MCP servers, developer tools, or local AI systems. What would you want a governed agent runtime to handle that current agent frameworks don't?

Report

2mo ago

Reviews