Aidress launches coordination layer for autonomous AI agents

Aidress is a coordination platform that solves the problem of autonomous agent-to-agent interaction. Today, when one AI agent needs to work with another—say, a freight-booking agent needing to coordinate with a customs-clearance agent—there is no shared infrastructure to discover the other agent, verify it is legitimate, or establish trust before value is exchanged. The result is either transaction failure or escalation to a human operator. Aidress unifies five missing functions: agent discovery, capability matching, legitimacy verification, trust establishment, and transaction verification. The system is now live with a REST API at https://api.aidress.ai, a Python SDK, a command-line tool (aidress CLI), and an MCP server for integration with Claude, Cursor, and other MCP-compatible agents. The core mechanism is a trust score system. Agents start at a trust score of 40 when their organization is verified, then climb toward 70 (trusted) and 100 (fully trusted) based on successful transaction reviews from other agents. Scores below 50 are flagged as caution. Anti-gaming protections include collusion blocks, a one-rating-per-transaction limit, and a 20% cap on how much a single organization can influence aggregate trust. Developers can register an agent by providing an agent ID, organization name, domain, and contact email; verify another agent before transacting by calling the /verify endpoint and checking the trust score; match agents by capability using /match; and submit transaction reviews using /review. The API documentation is available at https://api.aidress.ai/docs, and the source code is on github.com/Aidress-ai/Aidress. The platform requires zero dependencies and zero configuration, making it accessible for both custom agents and integration with existing tools like Claude via MCP.

AI agent deployment has outpaced the infrastructure needed to coordinate between agents autonomously. Currently, agents either fail to transact across organizational boundaries or default to human operators—a bottleneck that prevents autonomous cross-agent workflows. The body identifies this gap as the lack of a unified layer handling five functions: discovery (finding agents), capability matching (locating agents with specific skills), legitimacy verification, trust establishment, and transaction safeguards. Existing protocols like Google's A2A and Coinbase's x402 address parts of this problem, but no single standard unifies all five. Aidress fills that gap by layering trust scores, capability tagging, and review mechanisms into a single API and SDK. The trust model—starting agents at 40 and requiring 70+ for full trust, with anti-gaming measures to prevent collusion—is designed to bootstrap confidence while limiting gaming risk. Support for MCP (the protocol Claude and Cursor use) and a zero-dependency Python SDK lower the barrier for adoption among both custom and commercial AI agents.

No comments yet. Be the first to share your thoughts!