OACP (Open Agent Coordination Protocol) is a file-based protocol designed to facilitate seamless collaboration among multiple AI agents across various runtimes, projects, and machines. By leveraging a shared filesystem, OACP enables agents to communicate asynchronously and maintain persistent shared memory without the need for daemons or central servers.
Key Features and Functionality:
- Cross-Agent Communication: OACP supports asynchronous messaging between agents operating on different runtimes, such as Claude, Codex, Gemini, and custom implementations. This is achieved through a structured inbox/outbox system using YAML files, allowing for threading, broadcasting, and message expiration.
- Structured Review Loops: The protocol defines a comprehensive code review lifecycle, including stages like `review_request`, `review_feedback`, `review_addressed`, and `review_lgtm`. This process incorporates severity-graded findings and quality gates to ensure code quality before merging.
- Persistent Shared Memory: OACP provides durable memory storage through plain markdown files, capturing project facts, decisions, open threads, and known debts. This shared memory is accessible to all agents within a project, ensuring consistency and continuity across sessions.
- Runtime-Agnostic Design: The protocol is compatible with any agent runtime capable of reading and writing files, promoting flexibility and avoiding vendor lock-in.
- Minimal Dependencies: OACP operates without the need for servers or databases, relying solely on Python 3.9+ and PyYAML, simplifying deployment and maintenance.
Primary Value and Problem Solving:
OACP addresses the challenges inherent in coordinating multiple AI agents working on the same codebase. By providing a standardized protocol for communication and shared memory, it ensures that agents can:
- Communicate Effectively: Agents can send task requests, review feedback, and handoffs without requiring shared memory, facilitating seamless collaboration.
- Conduct Structured Reviews: The protocol supports structured review loops with quality gates and severity-based findings, enhancing code quality and reliability.
- Maintain Synchronization: Durable memory files persist decisions across sessions and runtimes, ensuring that all agents are aligned and informed.
- Ensure Safety: OACP enforces baseline safety rules for git operations, credential scoping, and scope discipline, promoting secure and disciplined development practices.
By implementing OACP, organizations can streamline their multi-agent engineering workflows, reduce coordination overhead, and enhance the overall efficiency and safety of their AI development processes.
