Open-source infrastructure for controlled, inspectable agent teams.
Build and run multi-agent systems in your own environment. Take them to production with YuanASI.
For companies → · Explore OMA → · Documentation →
Open Multi-Agent is the public open-source organization behind OMA, a TypeScript-native orchestration framework for building multi-agent systems that can be approved, traced, evaluated, and resumed. OMA is built by YuanASI, which provides commercial engineering and support for teams moving agent systems from scope to production.
OMA turns a goal into a task DAG at runtime—or executes an explicit graph when the workflow must stay fixed.
- Orchestrate: plan, route, and schedule work across specialized agents.
- Control: approve consequential actions, bound cost, and preserve execution evidence.
- Operate: trace, evaluate, checkpoint, and resume runs on cloud or local models.
- Own the environment: run on your infrastructure and credentials, including private and offline deployments.
Run the no-key Quick Start → · Browse the repository →
YuanASI works with teams that need more than an open-source library alone:
- Custom AI Agent Delivery: scenario design, implementation, evaluation, and production deployment.
- Multi-Agent System Integration: architecture, internal-system integration, private deployment, reliability tuning, and team handoff.
- Enterprise AI Advisory: use-case assessment, technical design, POCs, ROI analysis, and delivery roadmaps.
Commercial inquiries: co@yuanasi.com
OMA is already used in publicly verifiable security analysis, pull-request review, coding assistants, document workflows, and fully offline deployments.