Matei Zaharia, the Apache Spark creator, has released Omnigent as a fully open-source meta-harness that layers on top of existing tools such as Claude Code and Codex, letting users compose multi-agent workflows, apply governance policies, and move live sessions between laptop and phone with no cloud dependency.
Really excited to open source a new project: Omnigent, a meta-harness for AI agents.
It lets you build multi-agent coding and custom agents, sitting above Claude Code, Codex, Pi, and agent SDKs to let you compose them. It also adds live collaboration and rich control policies.
The GitHub repo shows a single setup step that launches a terminal UI plus references to web, native, and mobile interfaces, all executing on the user's own hardware.
Live collaboration across Slack, Teams, CLI, and web is described, yet no adoption numbers or real-world team usage data appear in the announcement materials.
Users praised Databricks and Matei Zaharia's open-sourced Omnigent Meta-Harness as a game changer because it supplies the missing orchestration layer for coordinating teams of specialized AI agents.
This is super exciting - I’ve been using Omnigent at Databricks for a while, and today we open-sourced it.
Omnigent is a meta-agent for orchestrating a swarm of agents.
Why do we need this? The best results no longer come from a single model running in a single harness.
I used to run the same task with Codex and Claude Code, then pick the better one. But the obvious thing is to let them collaborate, debate, and converge on something better. Omnigent makes this smooth.
The other feature I love is real-time collaboration. You can invite people into an Omnigent session to watch, steer, and send commands.
Multi-agent, multi-human collaboration is the future.
Omnigent was built by @matei_zaharia and a very lean team in just 6 weeks, working every day out of a Databricks war room, truly amazing.
Databricks AI really feels like a startup.
Really excited to open source a new project: Omnigent, a meta-harness for AI agents.
It lets you build multi-agent coding and custom agents, sitting above Claude Code, Codex, Pi, and agent SDKs to let you compose them. It also adds live collaboration and rich control policies.
Check out Omnigent, an open source harness that lets you use all the existing code harnesses (Claude Code, Codex, OpenCode, pi), collaborate and share sessions in many modalities (e.g. Slack/Teams, cli, webui), while having a fine grained security model that really tightens the control on what agents can do/not do. https://github.com/omnigent-ai/omnigent
There's a lot more coming, e.g. meta-optimization with @gepa_ai and programmatic tool calling as in RLM and MemEx. Read more in our blog: https://www.databricks.com/blog/introducing-omnigent-meta-harness-combine-control-and-share-your-agents or get started at https://omnigent.ai/
Omnigent can be run on your laptop or deployed to @Docker, @Railway and @flydotio, and can run agents in the cloud on @modal and @daytonaio. It works with any LLM provider too (coding agents' subscriptions, @OpenRouter, @databricks, etc).
Own the harness, own the agent orchestrators.
Great to see open-source work starting to enable it.
Being able to compose and combine multiple agents is clearly the future to avoid model lock-in. Curious how routing works, as that remains unsolved.
Really excited to open source a new project: Omnigent, a meta-harness for AI agents.
It lets you build multi-agent coding and custom agents, sitting above Claude Code, Codex, Pi, and agent SDKs to let you compose them. It also adds live collaboration and rich control policies.
Omnigent is based on the trend we saw with AI usage at Databricks and Neon: engineers were combining multiple agents into loops and workflows, and this was difficult above the harness layer. We add a uniform API above any harness that enables rich features on top.
Really excited to open source a new project: Omnigent, a meta-harness for AI agents.
It lets you build multi-agent coding and custom agents, sitting above Claude Code, Codex, Pi, and agent SDKs to let you compose them. It also adds live collaboration and rich control policies.
Omnigent focuses on three problems above the level of a single harness: composition, collaboration and control. https://www.databricks.com/blog/introducing-omnigent-meta-harness-combine-control-and-share-your-agents
For composition, it lets you create multi-agent teams with different harnesses or swap harness and model mid-session and mid-loop.
Omnigent is based on the trend we saw with AI usage at Databricks and Neon: engineers were combining multiple agents into loops and workflows, and this was difficult above the harness layer. We add a uniform API above any harness that enables rich features on top.
Incredibly excited about this launch! Imagine you lost connection to your agent's LLM partway through a session... Omnigent would detect that failure and delegate the work to a new agent. Unified interface w/ robust execution.
Really excited to open source a new project: Omnigent, a meta-harness for AI agents.
It lets you build multi-agent coding and custom agents, sitting above Claude Code, Codex, Pi, and agent SDKs to let you compose them. It also adds live collaboration and rich control policies.
@omarsar0 We have a few examples coming for difficulty-based routing but it is definitely a big open area.
Own the harness, own the agent orchestrators.
Great to see open-source work starting to enable it.
Being able to compose and combine multiple agents is clearly the future to avoid model lock-in. Curious how routing works, as that remains unsolved.
For collaboration, Omnigent adds a common UI above any agents, with native app, web and mobile UIs and live collaboration with teammates. No more copy-pasting text and screenshots between agents, Slack, and Docs all the time! You can even review code with the agent that wrote it.
Omnigent focuses on three problems above the level of a single harness: composition, collaboration and control. https://www.databricks.com/blog/introducing-omnigent-meta-harness-combine-control-and-share-your-agents
For composition, it lets you create multi-agent teams with different harnesses or swap harness and model mid-session and mid-loop.
@kubla Yeah, we'll see where things land, but why risk having the "wrong" harness when you can try the ones from the model builders? We tried to make it very easy to configure them with the same skills/tools/etc and talk across them.
Omnigent can be run on your laptop or deployed to @Docker, @Railway and @flydotio, and can run agents in the cloud on @modal and @daytonaio. It works with any LLM provider too (coding agents' subscriptions, @OpenRouter, @databricks, etc).
@manonginusa @Yuchenj_UW I admit agents did most of the work ;).. but also built by a great team of engineers here across AI, security, and UX.
@0xAkvn Omnigent is already deployed at scale inside Databricks, thousands of engineers are using it.
You can also launch agents on hosted sandbox like @modal.
This collaboration UI is pretty cool btw.
@parkersmith @matei_zaharia Thank you for reporting! Just made the package public as well: https://github.com/omnigent-ai/omnigent/pkgs/container/omnigent-server
Note this has a slight different name than our docs. Will fix that too.
``` docker pull http://ghcr.io/omnigent-ai/omnigent-server:v0.1.0 ```
For security & cost control, Omnigent introduces a *contextual* policy model that can track session state (e.g. ask for permission to git push if an agent downloaded an untrusted npm package, or ask you after every $100 of spend) and a strong OS sandbox from our security team.
@Yuchenj_UW How is it meant to be deployed at scale?
@matei_zaharia @Yuchenj_UW so sound like you are actually leading and doing some of the project implementation work? that's still surprising as a CTO is still directly involved in such a project. most CTOs never look at a single project, even they do, they just care about result or just 'lead'
@matei_zaharia Super cool! Really love this initiative. I will be going deep into it and showcasing it to my community in the coming days.
@NirantK @matei_zaharia The name tripped me too for a second — I went in expecting something closer to DSPy territory.
@manonginusa @matei_zaharia @matei_zaharia drives projects and writes code with the team, and deeply involved in products and research.
Databricks is a founder-led company.