# Directory

A distributed, peer-to-peer system for publishing, discovering, and verifying metadata about AI agents by their capabilities and dependencies.

Directory (dir) is a system for publishing, exchanging, and discovering information about AI agent records over a distributed peer-to-peer network. It follows the Open Agentic Schema Framework to describe agent capabilities in a hierarchical taxonomy, so requirements can be matched to agents that satisfy them.

Records can be cryptographically signed to support data integrity and provenance tracking, and can be linked to express relationships such as version history, collaborative partnerships between complementary agents, and dependency chains in composite workflows. Content-addressing and distributed hash tables support discovery and synchronization across the network without a central server. The project provides a command-line client, Go, Python, and JavaScript SDKs, GitHub Actions, and an MCP server for interacting with the system, and is written in Go under the Apache 2.0 license.

Website: https://agntcy.github.io/dir/