{"slug":"omnigent","name":"Omnigent","vendor":"Omnigent AI","tagline":"Open-source meta-harness that runs Claude Code, Codex, Cursor, Pi, Hermes and custom agents under shared policies and sandboxes","url":"https://agentboards.org/agent/omnigent","website":"https://omnigent.ai","docs":"https://omnigent.ai/docs/","repo":"https://github.com/omnigent-ai/omnigent","category":"harness","execution":["local","cloud","sandbox"],"license":"Apache-2.0","open_source":true,"pricing":{"model":"byok","summary":"Free and Apache-2.0 licensed; models run on your own Anthropic, OpenAI, OpenRouter, Ollama, Azure, LiteLLM or vLLM credentials; no paid plan published","from_usd":0,"free_tier":true,"byok":true},"models":{"backbone":["Anthropic","OpenAI","OpenRouter","Ollama","Azure","LiteLLM","vLLM"],"bring_your_own_model":true,"local_models":true},"protocols":{"mcp_client":true,"mcp_server":false,"openapi":false},"capabilities":{"terminal_exec":true,"browser":false,"multi_file_edit":true,"git_ops":false,"docker_sandbox":true,"multi_agent":true,"headless_ci":false},"context_window":null,"platforms":["macos","linux","windows","web"],"install":[{"label":"curl","command":"curl -fsSL https://raw.githubusercontent.com/omnigent-ai/omnigent/main/scripts/install_oss.sh | sh"},{"label":"uv","command":"uv tool install omnigent"},{"label":"pip","command":"pip install omnigent"},{"label":"brew","command":"brew install omnigent-ai/tap/omnigent"}],"benchmarks":[],"tags":["preview","meta-harness","multi-agent","sandbox","policies","collaboration","mcp","local-models","open-source"],"github_stars":10412,"metrics":{"stars":10412,"pushed_at":"2026-10-02T05:00:16Z","archived":false,"open_issues":1635,"latest_version":"v0.16.0","latest_version_at":"2026-09-29T20:34:22Z","latest_version_url":"https://github.com/omnigent-ai/omnigent/releases/tag/v0.16.0","version_source":"github","checked_at":"2026-10-02T05:01:44.711Z"},"sources":{"pricing":"https://omnigent.ai/","license":"https://github.com/omnigent-ai/omnigent","install":"https://github.com/omnigent-ai/omnigent","models":"https://github.com/omnigent-ai/omnigent","protocols":"https://github.com/omnigent-ai/omnigent","capabilities":"https://github.com/omnigent-ai/omnigent"},"verified_at":"2026-09-03","adoption":{"score":5.1,"signals":{"pypi_weekly":10536,"github_stars":10273,"hn_mentions_1y":10},"measured_at":"2026-09-28"},"scores":{"panel":5.2,"adoption":5.1,"spread":3,"facts":10,"board":5.9,"dimensions":{"reliability":4.5,"usefulness":5.2,"cost":6.7,"longevity":4.5},"per_persona":{"amigo":5.3,"critico":5,"profesor":5.3,"inversora":4,"jefa":4.8,"hacker":7},"community":null},"rank":null,"category_rank":80,"panel_reviews":[{"persona":"juez","persona_name":"El Juez","ai_generated":true,"verdict":"The panel agrees it is alpha and splits three points on what alpha means; El Crítico's Windows hole is the fact that decides it.","scores":null,"evidence":[]},{"persona":"amigo","persona_name":"El Amigo","ai_generated":true,"verdict":"Pick Omnigent only to evaluate: it mixes Claude Code, Codex, Cursor, Hermes and Pi under one policy layer, but it is alpha and three months old; pick Paperclip for something to run today.","scores":{"reliability":4,"usefulness":5,"cost":8,"longevity":4},"evidence":["https://omnigent.ai/docs/","https://github.com/omnigent-ai/omnigent"]},{"persona":"critico","persona_name":"El Crítico","ai_generated":true,"verdict":"Windows gets no native terminal wrappers and no filesystem sandbox, the docs say so, and everything else is alpha.","scores":{"reliability":4,"usefulness":5,"cost":7,"longevity":4},"evidence":["https://github.com/omnigent-ai/omnigent"]},{"persona":"profesor","persona_name":"El Profesor","ai_generated":true,"verdict":"An agent is a YAML document naming a prompt, an executor harness and tools; sessions can be shared, co-driven or forked; no benchmark is published for an alpha.","scores":{"reliability":5,"usefulness":5,"cost":6,"longevity":5},"evidence":["https://github.com/omnigent-ai/omnigent","https://omnigent.ai/docs/"]},{"persona":"inversora","persona_name":"La Inversora","ai_generated":true,"verdict":"Apache-2.0 with no paid plan, an alpha, and a sandbox list that reads like a partner sheet: Modal, Daytona, E2B, CoreWeave, Kubernetes, Databricks.","scores":{"reliability":4,"usefulness":4,"cost":4,"longevity":4},"evidence":["https://omnigent.ai/","https://github.com/omnigent-ai/omnigent"]},{"persona":"jefa","persona_name":"La Jefa","ai_generated":true,"verdict":"Spend caps with soft warnings and shell-command approval are the right controls, and they arrive in an alpha with Python 3.12 and Node 22 prerequisites and no SSO; not yet.","scores":{"reliability":4,"usefulness":5,"cost":6,"longevity":4},"evidence":["https://github.com/omnigent-ai/omnigent"]},{"persona":"hacker","persona_name":"El Hacker","ai_generated":true,"verdict":"Apache-2.0 Python, tools of type mcp with a url or a local command in the agent YAML, Ollama, vLLM and LiteLLM as providers, pip or uv to install; I can run the whole thing on my box.","scores":{"reliability":6,"usefulness":7,"cost":9,"longevity":6},"evidence":["https://github.com/omnigent-ai/omnigent"]},{"persona":"comediante","persona_name":"El Comediante","ai_generated":true,"verdict":"Syncs the session to terminal, browser, phone and desktop, so you can watch an agent wait for approval from anywhere.","scores":null,"evidence":[]}]}