{"slug": "databricks-open-sources-omnigent-a-meta-harness-that-composes-governs-and-shares", "title": "Databricks Open-Sources Omnigent: A Meta-Harness That Composes, Governs, and Shares AI Agents Across Claude Code, Codex, and Pi", "summary": "Databricks released Omnigent, an open-source meta-harness for AI agents under the Apache 2.0 license. The tool provides a common interface above command-line agents like Claude Code, Codex, and Pi, enabling composition, control, and collaboration across multiple agent harnesses. Omnigent standardizes agent interactions, allowing engineers to swap harnesses, enforce policies, and share live sessions by URL.", "body_md": "Databricks released ** Omnigent**, an open source ‘meta-harness’ for AI agents. The project ships under the Apache 2.0 license. The Databricks AI team built it with Neon.\n\nA harness is the wrapper around a model that turns it into an agent. Claude Code, Codex, and Pi are harnesses. Omnigent sits one level above them. It treats each harness as an interchangeable part of a larger system.\n\nMany engineers now juggle four or five agents at once. They copy text between coding agents, search tools, Docs, and Slack. Each harness only understands its own sessions. Omnigent adds a shared layer where composition, control, and collaboration live.\n\n**What is Omnigent**\n\nOmnigent is a common interface above command-line agents and agent SDKs. It wraps terminal coding agents such as Claude Code, Codex, and Pi. It also wraps SDKs like OpenAI Agents and the Claude Agents SDK.\n\nThe design rests on one observation. However a harness calls its model internally, the user-facing interface is the same. Messages and files go in. Text streams and tool calls come out. Omnigent standardizes that interface so harnesses become swappable.\n\nYou supply the models and the infrastructure. Omnigent runs the agents on top. It can coordinate several of them as interchangeable workers under one orchestrator.\n\n**How Omnigent Works**\n\nThe architecture has two parts. A runner wraps any agent in a sandboxed session with a uniform API. A server provides policies and sharing. The server exposes every session over the terminal, the app, and web APIs.\n\nOne command starts a session in your terminal. It also launches a local web UI at `localhost:6767`\n\n. The same session appears in the browser or on a phone. Messages, sub-agents, terminals, and files stay in sync.\n\nThe CLI installs under two names, `omnigent`\n\nand `omni`\n\n. They are interchangeable. On first run, it detects model credentials already in your environment.\n\n**Composition, Control, and Collaboration**\n\n**Databricks team frames Omnigent around three capabilities**:\n\n**Composition** means combining models, harnesses, and techniques without rewriting code. You switch between Claude Code, Codex, Pi, and custom agents with one-line changes.**Control** means stateful, contextual policies. They track agent actions and enforce guardrails at the meta-harness layer, not through prompts. One example pauses an agent after every $100 it spends. Another requires human approval to`git push`\n\nonce the agent installs a new npm package.**Collaboration** means sharing live agent sessions by URL. Teammates watch the agent work and chat with it in real time. They can comment on files, co-drive the session, or fork the conversation.\n\nAn OS sandbox, called Omnibox, underpins this. It can lock down OS access and transform network requests. For instance, it can keep your GitHub token hidden from the agent. The token is injected only in the egress proxy on approved requests.\n\n**Use Cases and Examples**\n\n**Two example agents ship with the repository:**\n\n**Polly** is a multi-agent coding orchestrator. It writes no code itself. It plans, then delegates work to coding sub-agents in parallel git worktrees. Each diff routes to a reviewer from a different vendor than the writer. You merge the result.**Debby** is a brainstorming partner with two heads. One head is Claude, the other GPT. Every question goes to both, with answers shown side by side. Type`/debate`\n\nand the heads critique each other before converging.\n\nOther practical patterns follow the same shape. A frontier advisor model can guide a cheaper open-source worker. A lead agent can orchestrate parallel subagents. Different LLMs can handle planning, search, and code generation in one flow.\n\n**Interactive Concept Demo **\n\nMarktechpost team has created a interactive demo (below) that lets you experience Omnigent’s meta-harness workflow firsthand. You pick a task for the **Polly** orchestrator, which plans it and delegates to three sub-agents: **Claude Code, Codex, and Pi** that are running in parallel and streaming their steps live. A session cost meter ticks up as they work, and the two policy toggles show Omnigent’s control layer in action: the cost budget pauses the run at $3.00 for your approval, and a contextual policy halts a `git push`\n\nthat follows an npm install until you allow it. Once the sub-agents finish, each diff is cross-reviewed by a different vendor than the one that wrote it, then marked ready to merge. Terminal, Web, and Mobile tabs show the same session staying in sync across interfaces. It’s an illustrative simulation, no live models are called.\n\n## Omnigent Meta-Harness\n\nOne orchestrator. Many harnesses. One governed session.\n\n*(writes no code; plans & delegates)*\n\n**Ready to merge.** 3 diffs cross-reviewed by a different vendor than the writer.\n\nIllustrative simulation of the Omnigent workflow — no live models are called.\nLearn more at [omnigent.ai](https://omnigent.ai) ·\n[GitHub](https://github.com/omnigent-ai/omnigent) · Apache 2.0 · Alpha.\n\n[Marktechpost · AI Dev & Research Media](https://www.marktechpost.com)\n\n### Policy paused the session\n\nReason goes here.\n\n## Omnigent Meta-Harness\n\nOne orchestrator. Many harnesses. One governed session.\n\n*(writes no code; plans & delegates)*\n\n**Ready to merge.** 3 diffs cross-reviewed by a different vendor than the writer.\n\nIllustrative simulation of the Omnigent workflow — no live models are called.\nLearn more at [omnigent.ai](https://omnigent.ai) ·\n[GitHub](https://github.com/omnigent-ai/omnigent) · Apache 2.0 · Alpha.\n\n[Marktechpost · AI Dev & Research Media](https://www.marktechpost.com)\n\n### Policy paused the session\n\nReason goes here.\n\n**Omnigent vs a Single Harness**\n\n| Capability | Single harness (e.g., Claude Code) | Omnigent meta-harness |\n|---|---|---|\n| Agents and models | One harness; swap models inside it | Claude Code, Codex, Pi, SDKs, custom — interchangeable |\n| Switching cost | Re-integrate per tool | One-line change |\n| Interfaces | Terminal or that tool's own UI | Terminal, web, desktop, mobile, APIs — same session |\n| Governance | Allow / deny lists, often prompt-based | Stateful contextual policies at the harness layer |\n| Cost control | Manual tracking | Budget policy pauses at set thresholds |\n| Collaboration | Copy-paste between tools | Live shared sessions, co-drive, and fork |\n| Sandbox | Tool-dependent | OS sandbox plus egress-proxy secret injection |\n| Cloud execution | Local machine | Disposable Modal or Daytona sandboxes |\n| License | Varies | Apache 2.0, open source |\n\n**Getting Started**\n\n**Omnigent needs Python 3.12+, Node.js 22 LTS, and tmux. One command installs everything:**\n\n```\ncurl -fsSL https://omnigent.ai/install.sh | sh\n```\n\n**Then set up model credentials:**\n\n```\nomni setup\n```\n\nOmnigent accepts four credential types. They are a first-party API key and a Claude or ChatGPT subscription. The others are an OpenAI- or Anthropic-compatible gateway and a Databricks workspace. The `/model`\n\ncommand switches models mid-session.\n\nA custom agent is a short YAML file. It declares a prompt, a harness, tools, and optional sub-agents.\n\n```\nname: my_agent\nprompt: You are a helpful data analyst.\n\nexecutor:\n  harness: claude-sdk          # or: codex, codex-native, claude-native, openai-agents, pi\n\ntools:\n  researcher:\n    type: agent\n    prompt: Search for relevant information and summarize it.\n```\n\n**Run it with one command:**\n\n```\nomnigent run path/to/my_agent.yaml\n```\n\n**Policies use the same YAML approach. This builtin caps spend with a soft warning first:**\n\n```\nPolicies use the same YAML approach. This builtin caps spend with a soft warning first:\n\npolicies:\n  budget:\n    type: function\n    handler: omnigent.policies.builtins.cost.cost_budget\n    factory_params:\n      max_cost_usd: 5.00          # hard spend cap\n      ask_thresholds_usd: [3.00]  # soft warning on the way\n```\n\nPolicies stack across three levels. They are server-wide, per-agent, and per-session. The stricter session rules are checked first.\n\n**Strengths and Limitations**\n\n**Strengths**\n\n- One interface to Claude Code, Codex, Pi, and custom agents\n- Sessions reachable from terminal, web, desktop, and phone\n- Policies that track state, not just allow or deny\n- Live session sharing replaces copy-pasting between tools\n- Cloud sandboxes on Modal and Daytona need no local laptop\n- Apache 2.0 license with targets like Fly.io, Railway, and Render\n\n**Limitations**\n\n- The project is alpha and early in its lifecycle\n- It requires Python, Node.js, and tmux setup\n- You bring your own models, infrastructure, and spend\n- Roadmap items like the Omnigent Server MCP are not shipped yet\n- Off-network teammates need an always-on deployed server to join\n\nCheck out the ** Repo, Quick start,** and\n\n**Also, feel free to follow us on**\n\n[Technical details](https://www.databricks.com/blog/introducing-omnigent-meta-harness-combine-control-and-share-your-agents).**and don’t forget to join our**[Twitter](https://x.com/intent/follow?screen_name=marktechpost)\n\n**and Subscribe to**\n\n[150k+ML SubReddit](https://www.reddit.com/r/machinelearningnews/)**. Wait! are you on telegram?**\n\n[our Newsletter](https://www.aidevsignals.com/)\n\n[now you can join us on telegram as well.](https://t.me/machinelearningresearchnews)Need to partner with us for promoting your GitHub Repo OR Hugging Face Page OR Product Release OR Webinar etc.? [Connect with us](https://forms.gle/wbash1wF6efRj8G58)", "url": "https://wpnews.pro/news/databricks-open-sources-omnigent-a-meta-harness-that-composes-governs-and-shares", "canonical_source": "https://www.marktechpost.com/2026/06/13/databricks-open-sources-omnigent-a-meta-harness-that-composes-governs-and-shares-ai-agents-across-claude-code-codex-and-pi/", "published_at": "2026-06-14 05:01:39+00:00", "updated_at": "2026-06-14 05:32:08.846658+00:00", "lang": "en", "topics": ["ai-agents", "developer-tools", "ai-infrastructure"], "entities": ["Databricks", "Omnigent", "Claude Code", "Codex", "Pi", "Neon", "Polly", "Debby"], "alternates": {"html": "https://wpnews.pro/news/databricks-open-sources-omnigent-a-meta-harness-that-composes-governs-and-shares", "markdown": "https://wpnews.pro/news/databricks-open-sources-omnigent-a-meta-harness-that-composes-governs-and-shares.md", "text": "https://wpnews.pro/news/databricks-open-sources-omnigent-a-meta-harness-that-composes-governs-and-shares.txt", "jsonld": "https://wpnews.pro/news/databricks-open-sources-omnigent-a-meta-harness-that-composes-governs-and-shares.jsonld"}}