Repolode
Submit
omnigent-ai

omnigent-ai/omnigent

Open-source meta-harness that orchestrates Claude Code, Codex, Cursor, and custom agents with shared policies, sandboxing, and real-time collaboration.

View on GitHubWebsite
Stars
11,273
Forks
1,867
Open issues
2,396
Last commit
2 hours ago

What is it?

Omnigent is an open-source AI agent framework that sits above existing coding agents. It provides one orchestration layer over Claude Code, Codex, Cursor, OpenCode, Hermes, Pi, and agents you define yourself in YAML, so you can swap or combine harnesses without rewriting your workflow. Sessions can mix several agents, for example one agent reviewing another's work or a task split across agents.

The project covers governance and access. Policies can pause for approval before risky actions, cap spending, or limit which tools an agent can reach, and they apply at the server, agent, or session level. Agents can run in local or cloud sandboxes, including Modal, Daytona, E2B, Kubernetes, and Databricks. Model access supports API keys, Claude or ChatGPT subscriptions, and OpenAI- or Anthropic-compatible gateways.

Sessions can be started in a terminal and continued in a browser or on a phone through the web UI, and a native macOS desktop app is also available. Teammates can watch a shared session live, co-drive it, or fork a conversation. The project is installed with Python 3.12+ and is currently labeled alpha.

Key features

  • 01Runs Claude Code, Codex, Cursor, OpenCode, Hermes, Pi, and custom YAML agents in one session.
  • 02Applies approval, spend-cap, and tool-access policies at server, agent, or session level.
  • 03Runs agents in local or cloud sandboxes such as Modal, E2B, Daytona, and Kubernetes.
  • 04Supports API keys, Claude or ChatGPT subscriptions, and compatible model gateways.
  • 05Syncs sessions across terminal, browser, phone, and desktop app.
  • 06Lets teammates share, co-drive, or fork live sessions.

Use cases

  • Comparing or combining several coding agents on the same task.
  • Letting an agent run in a cloud sandbox while you monitor it from your phone.
  • Requiring human approval before agents run shell commands or spend budget.
  • Pairing with a teammate on a live agent session.

Quick start

  1. 1

    Install Omnigent with uv (Python 3.12+ required)

    uv tool install omnigent
  2. 2

    Start a session in your terminal

    omnigent
  3. 3

    Open the local web UI at http://localhost:6767 in your browser

Things to know

  • The project is marked alpha, so expect breaking changes.
  • Windows support is degraded; several terminal wrappers and sandboxing features are unavailable there.
  • Linux needs bubblewrap and tmux for the native harness terminals, and the harness CLIs need Node.js 22 or newer.

In collections

Multi-agent orchestration frameworksClaude Code tools, skills & pluginsOpen-source AI agent frameworks & appsAI coding assistants & agents

Topics

#agent-framework#ai-agents#coding-agents#multi-agent#sandbox#llm#claude-code#agent-governance#agent-orchestration#agents

FAQ

What is omnigent?
Open-source meta-harness that orchestrates Claude Code, Codex, Cursor, and custom agents with shared policies, sandboxing, and real-time collaboration. Omnigent is an open-source AI agent framework that sits above existing coding agents. It provides one orchestration layer over Claude Code, Codex, Cursor, OpenCode, Hermes, Pi, and agents you define yourself in YAML, so you can swap or combine harnesses without rewriting your workflow. Sessions can mix several agents, for example one agent reviewing another's work or a task split across agents.
Is omnigent free to use? What license does it have?
omnigent is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
Is omnigent still maintained?
Actively maintained. The last commit to omnigent was on Nov 7, 2026.
What can I use omnigent for?
Comparing or combining several coding agents on the same task. Letting an agent run in a cloud sandbox while you monitor it from your phone. Requiring human approval before agents run shell commands or spend budget. Pairing with a teammate on a live agent session.
How do I install omnigent?
Install Omnigent with uv (Python 3.12+ required): uv tool install omnigent Start a session in your terminal: omnigent Open the local web UI at http://localhost:6767 in your browser
Where can I get omnigent?
The source code and installation instructions are on GitHub: https://github.com/omnigent-ai/omnigent

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects