Open Multi-Agent

Open Multi-Agent (OMA) is a self-hosted TypeScript framework for building and running multi-agent systems that your organization can own and audit. It enables teams of AI agents to work in parallel on complex tasks, with consequential actions requiring durable, tamper-evident human approvals before execution. Every run creates an offline-verifiable record, ensuring transparency and accountability.

Designed for enterprises and developers needing control over agent behavior, OMA supports multiple model providers (OpenAI, local servers, Chinese providers) and includes built-in approval gates, run journaling, and offline verification. The project describes itself as built with TypeScript and includes tools for code generation via Claude.