open-multi-agent

A self-hosted TypeScript runtime for running AI agents in an organization. Risky actions wait for approval, and every run leaves a verifiable record for auditing.

Share on XLicense: MIT

Overview

OMA (Open Multi-Agent) is a self-hosted TypeScript agent runtime for organizations that want to own and audit their AI agents. Consequential actions wait for durable, tamper-evident approvals, and every run leaves a record that can be verified offline. It has no telemetry and no hosted control plane, and works with cloud models, local ones such as Ollama and vLLM, or Chinese providers.

Key features

  • Durable, tamper-evident approvals for risky actions
  • Run records verifiable offline
  • No telemetry and no hosted control plane
  • Cloud, local and Chinese model providers
  • Starter templates via npm create oma-app

Best for

Teams that need governance, approvals and audit trails around AI agents in their own environment. It requires Node.js 20 or newer, and a maintained LTS release is advised for production.

Upstream
open-multi-agent/open-multi-agent
Fork on GitHub
Guo-astro/open-multi-agent
Upstream stars
7.0k
Category
AI agents and LLM tools
Language
TypeScript
License
MIT
Forked
2026-04-04
Sync status
In syncLast synced 2026-09-29