Skip to main content
AdOpenFree logoPromote your productReach more potential users and drive product growth and revenue.Advertise
Favicon of Omnigent

Omnigent

Free Listing

Open-source meta-harness for AI agents: a common layer over Claude Code, Codex, Cursor, Pi, and custom agents. Swap harnesses without rewriting, enforce policies and sandboxing, collaborate in real time from any device.

Visit omnigent-ai/omnigent

Overview

Omnigent is an open-source meta-harness that provides a common orchestration layer over AI agent harnesses like Claude Code, Codex, Cursor, and Pi, as well as custom agents. It enables swapping or combining harnesses without rewriting code, enforces policies and sandboxing, and supports real-time collaboration on live sessions from any device.

Key Features

  • Built-in multi-AI agents: Polly (coding orchestrator) and Debby (model debate), plus custom agents defined in YAML.
  • Contextual policies: stateful spend caps, model routing, and risk-based escalation.
  • Secure OS sandbox: restricts filesystem and network access, hides credentials, enables safe YOLO mode.
  • Composition: combine multiple models, harnesses, and techniques without rewriting code; switch between Claude Code, Codex, Pi, and custom agents with one-line changes.
  • Collaboration: share live agent sessions via URL with full history, allowing teammates to review, comment, and steer in real time.
  • Cloud sandboxes: run agents in disposable cloud sandboxes from providers like Modal, Daytona, and others.

Use Cases

  • Orchestrating multiple coding agents (e.g., Polly delegating to Claude Code, Codex, Pi) in parallel git worktrees.
  • Running model debates (Debby) to compare answers from different models side by side.
  • Collaborative debugging or investigation with teammates sharing a live session and co-driving on the same machine.
  • Governing agent actions with policies for cost caps, tool access limits, and approval workflows.

Getting Started

  • Install with `curl -fsSL https://omnigent\.ai/install\.sh | sh` or `uv tool install omnigent`. Run `omnigent` to start a session, `omnigent setup` to configure models, and `omnigent run examples/polly/` to try built-in agents.

Deployment & Requirements

  • Requires Python 3.12+; Node.js 22 LTS and npm/pnpm for coding-harness CLIs and web UI; tmux for native terminal wrappers; bubblewrap on Linux for sandboxing. For server deployment, use docker compose or one-click deploys on Render, Railway, Fly.io, etc.

Before You Adopt

  • License: Apache-2.0. Review its terms before using, modifying, or distributing the project.
  • Omnigent is in alpha stage and developed openly.
  • Telemetry is collected by default, but can be opted out.
  • Windows support is degraded; some features are unavailable.

Comments

Sign in to leave a comment.

More like Omnigent

Favicon of Poki

Poki

Free ListingStars: 0.5K

Let AI agents operate a real iPhone

Agent Frameworks

Poki is an open-source local MCP server and agent that lets AI agents operate a real iPhone from a Mac without jailbreak or extra hardware. Verified runs compile into routes so similar tasks take a shorter path.

Favicon of Rivet

Rivet

Free ListingStars: 6.2K

The orchestrator for agentic workloads.

Agent Frameworks

Rivet is a fast, high-density, scalable orchestrator for agentic workloads. It runs agents, workflows, and sandboxes as durable Actors, providing stateful primitives for AI agents, collaborative apps, and durable execution.

Favicon of sapper-v2

sapper-v2

Free ListingStars: 0.9K

AI commander and worker swarm for autonomous pentests

Agent Frameworks

sapper-v2 is an autonomous pentest legion: an AI commander coordinates general, scout, and soldier agents through a shared layered fact-intent graph, using model tiering and tmux sessions for unattended security testing campaigns.