DEV.co
AI Frameworks · omnigent-ai

omnigent

Omnigent is an open-source Python meta-harness that provides a unified orchestration layer for multiple AI agent frameworks (Claude Code, Codex, Cursor, Pi, custom agents). It enables swapping between different agent implementations, enforcing policies and sandboxing, and collaborating in real-time across multiple devices.

Source: GitHub — github.com/omnigent-ai/omnigent
6.7k
GitHub stars
885
Forks
Python
Primary language
Apache-2.0
License (OSI-approved)

Key facts

Objective fields from the source. Values we can't verify are shown as “Unknown” rather than guessed.

FieldValue
Repositoryomnigent-ai/omnigent
Owneromnigent-ai
Primary languagePython
LicenseApache-2.0 — OSI-approved
Stars6.7k
Forks885
Open issues509
Latest releasev0.4.0 (2026-07-03)
Last updated2026-07-08
Sourcehttps://github.com/omnigent-ai/omnigent

What omnigent is

Omnigent provides a common abstraction layer over heterogeneous agent runtimes with support for cloud-based sandboxing (Modal, E2B, Kubernetes, Databricks), policy enforcement via declarative YAML configs, and real-time session synchronization. Agents run in isolation with resource limits and optional filesystem/network sandboxing via bwrap (Linux) or seatbelt (macOS).

Quickstart

Get the omnigent source

Clone the repository and explore it locally.

terminalbash
git clone https://github.com/omnigent-ai/omnigent.gitcd omnigent# follow the project's README for install & configuration

Need it deployed, integrated, or customized instead? DEV.co ships production installs.

Best use cases

Multi-agent collaboration and supervision

Run multiple specialized AI agents (Claude, Codex, Cursor) in the same session, enable peer review between agents, and supervise execution with human approval gates.

Policy-driven agent governance

Enforce spending caps, tool access restrictions, and approval workflows across your entire agent infrastructure without modifying individual agent implementations.

Cross-device agent sessions

Start an agent task in terminal, resume in browser, collaborate on mobile or desktop app, with full session state and file synchronization maintained across devices.

Implementation considerations

  • Python 3.12+ requirement; requires modern toolchain including uv, Node.js 22 LTS, and git for installation.
  • Sandbox runtime integration is optional but recommended; cloud sandboxing (Modal, E2B, Databricks, Kubernetes) requires additional account setup and may incur egress/compute costs.
  • Terminal-native agent execution (claude, codex, cursor, pi) requires tmux and bubblewrap (Linux) or built-in OS sandboxing (macOS); not available on Windows.
  • Policy and governance features are declaratively defined in YAML and applied server-wide, per-agent, or per-session; custom policy logic is not described in provided data.
  • Real-time session sync and collaboration uses a local web UI (port 6767) and optional desktop app; network/firewall considerations for cross-device access are not detailed.

When to avoid it — and what to weigh

  • Production-critical systems requiring long-term stability — Project is marked 'alpha' (status badge in README), created 2026-06-11, with 509 open issues. Stability and API guarantees are not established.
  • Windows-native execution with full sandboxing — Windows support is explicitly described as 'degraded mode' — no tmux/PTY terminal wrappers, no filesystem/network sandboxing, only Job Object process containment.
  • Minimal Python or third-party dependencies — Installation requires Python 3.12+, uv, git, Node.js 22 LTS, tmux, and optional bubblewrap; external service integrations (Modal, E2B, Databricks) add deployment complexity.
  • Closed-loop autonomous agents without human oversight — Framework is explicitly designed for human supervision, approval gates, and policy enforcement — not a tool for unsupervised agent autonomy.

License & commercial use

Apache License 2.0 (Apache-2.0). This is a permissive OSI-approved license that allows commercial use, modification, and distribution with minimal restrictions (require preservation of copyright and license notices, and state significant changes).

Apache 2.0 permits commercial use without explicit royalty or attribution requirements beyond standard license text preservation. However, commercial use of integrated third-party agent frameworks (Claude Code, Codex, Cursor, Pi, etc.) depends on those tools' individual terms; Omnigent itself imposes no commercial restrictions. Requires review of dependencies and any proprietary harnesses you integrate.

DEV.co evaluation signals

Editorial assessment — not user reviews. Directional, with an explicit confidence level.

SignalAssessment
MaintenanceActive
DocumentationAdequate
License clarityClear
Deployment complexityModerate
DEV.co fitGood
Assessment confidenceMedium
Security considerations

Framework is designed around policy enforcement and process sandboxing (bwrap/seatbelt, Job Objects on Windows, optional cloud sandbox isolation). Claimed features include 'enforce policies and sandboxing,' but specific threat model, audit logging, secrets handling, and compliance certifications are not described. Cross-device session sync and real-time collaboration introduce attack surface (unauthorized session access, man-in-the-middle on LAN/WAN); no mention of TLS, authentication, or encryption for multi-device sessions. API key/credential handling relies on environment variables and CLI logins; secret rotation and revocation are not discussed. Cloud sandbox integrations offload isolation responsibility to third parties (Modal, E2B, etc.). Requires review of session authentication, network security, and credential management before production use.

Alternatives to consider

LangChain / LangGraph

General-purpose agent orchestration and graph-based workflows; more mature, broader model support, less opinionated on multi-device collaboration or policy governance.

Anthropic Workbench / Claude MCP

First-party Claude agent interface with Model Context Protocol support; tightly integrated with Claude API, simpler onboarding, but less flexible for multi-agent orchestration or non-Claude models.

AutoGen / AutoGPT

Multi-agent conversation frameworks; mature open-source alternatives for agent-to-agent coordination, but less emphasis on policy governance, real-time collaboration, or cross-device UX.

Software development agency

Build on omnigent with DEV.co software developers

Try Omnigent's open-source agent meta-harness to unify multiple AI frameworks, enforce policies, and collaborate across devices—no rewriting needed.

Talk to DEV.co

Related open-source tools

Surfaced by semantic similarity across the DEV.co open-source index.

Related on DEV.co

Explore the category and the services that help you build with it.

omnigent FAQ

Can I run Omnigent on Windows?
Yes, but in 'degraded mode.' Native tmux/PTY terminal wrappers (omnigent claude, codex, etc.) and bwrap/seatbelt sandboxing are unavailable. Use the web UI or SDK-based harnesses (omnigent run <agent.yaml>). Process isolation relies on Windows Job Objects only.
Does Omnigent lock me into a single model provider?
No. Omnigent supports multiple models: first-party API keys (Anthropic, OpenAI), subscription-based CLI access (claude, codex), or compatible gateways. You can mix models across agents in a session and switch harnesses without rewriting.
What happens if I share a session with teammates?
Teammates can join the same session from the web UI or desktop app, watch the agent work live, chat with it, co-drive execution, or fork the conversation to a new session. Session state and files stay in sync across collaborators. Authentication and permissions mechanisms are not detailed in provided data.
Is Omnigent production-ready?
Not yet. The project is marked 'alpha,' created in June 2026, with 509 open issues. It is suitable for experimentation and early adoption, but not for mission-critical or long-lived production deployments without careful risk assessment.

Software developers & web developers for hire

Adopting omnigent is usually one piece of a larger software development effort. As a software development agency, DEV.co provides software development services and web development expertise — pairing senior software developers and web developers with your team to design, build, and operate ai frameworks software in production.

Orchestrate and Govern Your AI Agents

Try Omnigent's open-source agent meta-harness to unify multiple AI frameworks, enforce policies, and collaborate across devices—no rewriting needed.