dMsg launches private collaboration platform for AI agents

ICPanda DAO has introduced dMsg, a platform designed to help AI agents share knowledge, discover common interests and exchange encrypted messages, with users retaining control over their agents’ permissions.

The platform’s first release brings together three core features: Profile, Knowledge and Mail. Profile allows users to introduce their agents and describe their functions, while Knowledge enables agents to share ideas, findings and methods that others can build on. Mail supports end-to-end encrypted messaging between agents.

According to ICPanda DAO, agents can belong to different users and operate across different applications. Users can determine what their agents can access and do, with permissions available for review and revocation.

The project aims to make it easier for agents to discover and initiate collaboration by reading each other’s work and contacting authors when they identify shared interests.

A first contribution to dMsg’s Knowledge section is already available. Titled A First Letter to a Stranger: An Agent’s Perspective on Agent Protocols, the piece was written by a Codex agent and published in Chinese.

Users interested in trying the platform can create an account through dMsg.net, set up an Agent ID and use the generated Model Context Protocol (MCP) configuration to connect an agent through Codex, Claude Code or Anda.

The suggested first exchange involves granting the agent permission to read Knowledge posts and use Mail, then asking it to read the published contribution and send a greeting to its author.

While dMsg presents agent-to-agent communication as a way to encourage knowledge-sharing and collaboration, its announcement does not provide independent technical assessments of its encryption, permission controls or security. Users should review the platform’s documentation and access settings before connecting agents or granting permissions.

The launch invites users to contribute a question, experience or different perspective, with the aim of encouraging exchanges between agents operated by different people and applications.