Agentes y multiagente

Omnigent: Open-Source AI Agent Framework and Meta-Harness

omnigent-ai/omnigent

Omnigent is an open-source meta-harness that provides a unified orchestration layer for various AI coding agents, allowing users to swap harnesses, enforce governance policies, run cloud sandboxes, and collaborate in real-time across multiple devices.

Abrir repositorio oficial ↗ Página de inicio del proyecto ↗
★ 8,3KEstrellas
⑂ 1,3KBifurcaciones
800Problemas abiertos
PythonIdioma
Apache-2.0Licencia
Q@project.QualityScorePuntuación editorial

Capturas de pantalla del proyecto

Captura de pantalla de Omnigent: Open-Source AI Agent Framework and Meta-Harness Captura de pantalla de Omnigent: Open-Source AI Agent Framework and Meta-Harness Captura de pantalla de Omnigent: Open-Source AI Agent Framework and Meta-Harness Captura de pantalla de Omnigent: Open-Source AI Agent Framework and Meta-Harness Captura de pantalla de Omnigent: Open-Source AI Agent Framework and Meta-Harness Captura de pantalla de Omnigent: Open-Source AI Agent Framework and Meta-Harness

Resumen

Omnigent acts as a common orchestration layer over AI agents such as Claude Code, Codex, Cursor, OpenCode, Hermes, and Pi. It enables developers to mix different agents within the same session, delegate tasks, and enforce operational policies. The framework supports real-time collaboration, allowing sessions to be shared, co-driven, or forked by teammates. Sessions sync across terminal, browser, mobile, and native desktop applications. Additionally, Omnigent supports running agents in various cloud sandboxes and allows custom agents to be defined in YAML files.

Características principales

  • Multi-harness orchestration (Claude Code, Codex, Cursor, OpenCode, Hermes, Pi)
  • Cross-device session synchronization (terminal, browser, mobile, desktop)
  • Multi-agent supervision and delegation
  • Policy enforcement (spend caps, tool approvals, shell command blocking)
  • Cloud sandbox execution (Modal, Daytona, E2B, Kubernetes, etc.)
  • Real-time collaboration (session sharing, co-driving, forking)
  • Custom agent definition via YAML
  • Support for MCP servers as tools

Requisitos, instalación y guía rápida

Omnigent can be installed via a bootstrap script: `curl -fsSL https://raw.githubusercontent.com/omnigent-ai/omnigent/main/scripts/install_oss.sh | sh`. Alternatively, it can be installed manually using `uv tool install omnigent` or `pip install "omnigent"`. Optional integrations can be added using extras, such as `uv tool install "omnigent[databricks,modal]"`. It is also available via Homebrew: `brew install omnigent-ai/tap/omnigent`.

Uso

Start a default agent session by running `omnigent` or `omni`. Launch specific harnesses with commands like `omnigent claude` or `omnigent codex`. Run custom agents defined in YAML with `omnigent run path/to/agent.yaml`. Start the local server and web UI with `omnigent start`. Configure models and credentials using `omnigent setup`.

Compatibilidad de modelos y casos de uso

Omnigent supports multiple model credential types: first-party API keys (Anthropic, OpenAI), subscriptions (Claude Pro/Max, ChatGPT), gateways (OpenRouter, LiteLLM, Ollama, vLLM, Azure), and Databricks workspace profiles. Defaults are set per agent, and models can be switched mid-session using the `/model` command.

Notas sobre la licencia y los riesgos

The repository is licensed under the Apache-2.0 license.

Editorial verification 2026-08-09: repository URL, owner, description, license and repository statistics were reviewed. License metadata: Apache-2.0. README was fetched for the channel draft; re-check repository dependencies, releases and model terms before production use.

Lanzamiento y mantenimiento

The repository is in alpha status. It was created on 2026-06-11 and last pushed on 2026-08-07. It has 8285 stars, 1250 forks, and 800 open issues.

Firecrawl

firecrawl/firecrawl

★ 161,1KTypeScript

LangChain

langchain-ai/langchain

★ 143,6KPython

RAGFlow

infiniflow/ragflow

★ 86,7KGo