Meta Buys Conductor to Run Teams of Coding Agents in Parallel

Meta buys the orchestra, not the musicians

The Meta Conductor acquisition was confirmed on November 4, 2026. Meta has bought Conductor, a startup founded in 2024 that lets developers run several AI coding agents at the same time, each in its own isolated workspace. Axios broke the news, TechCrunch confirmed it with Meta, and founder Charlie Holtz announced it the same day. Financial terms were not disclosed.

Conductor does not build its own AI models. That is the interesting part. Its platform coordinates coding agents from other companies, including Anthropic and OpenAI, giving each agent a separate copy of the codebase so several of them can work on different tasks in parallel instead of queueing for one shared session. The developer reviews the results and merges the good ones. Conductor is the coordination layer, not the brain, and Meta just bought the coordination layer.

What Conductor actually does

The problem Conductor solves is easy to recognize if you have used an AI coding agent. One agent working on one task is useful. Two agents working on the same codebase at the same time is a mess, because they overwrite each other's changes and lose track of who did what. Conductor gives each agent its own isolated workspace, a clean copy of the code, so they can work in parallel and the developer stays in charge of merging.

The product launched in July 2025, led by CEO Charlie Holtz and CTO Jackson de Campos. In March 2026 it raised a $22 million Series A from Spark and Matrix, with Y Combinator participating. Its newer Conductor Cloud moved agents into isolated virtual machines that can keep working for hours, which is the direction the whole field is moving: agents that run in the background while you do something else.

The team now joins Meta Superintelligence Labs, where Holtz and de Campos will work with chief AI officer Alexandr Wang and Nat Friedman, who leads product there. In August 2026 Meta launched Muse Code, its own AI coding agent, and Conductor's orchestration fits directly into that effort. Meta's agents are getting capable. The question is how to run a team of them without chaos, and that is what Conductor knows how to do.

If you use Conductor, mind the dates

The acquisition comes with a shutdown schedule that existing users need to read carefully. New sign-ups and downloads stopped on November 4. Conductor Cloud workspaces become read-only on November 13, which means you can still view old chats and export data, but you cannot send new messages. The cloud service shuts down completely on November 20, after which Conductor says it will begin deleting cloud user data.

The local desktop app keeps working until early 2027, when it stops getting new models and critical fixes. Holtz has promised something new early next year, which presumably means a Meta product built on Conductor's technology. For now, the practical advice is simple: if you have work in Conductor Cloud, export it before November 20, because deletion after that date is not a threat, it is the plan. The company also says it has never sold customer data or used it to train AI models, and it plans to delete user data when the existing application shuts down.

The same day, the same idea, twice

November 4 was a busy day for developer-tool acquisitions. Hours apart, OpenAI announced plans to acquire Pierre Computer Company, the maker of Code Storage and open-source tools called Trees, Diffs and Highlights, to bolster Codex. We covered that deal separately in our report on OpenAI's Pierre acquisition, and the parallel is hard to miss.

Both deals are about the same thing: the model is no longer the product. The product is the system around the model, the file navigation, the change review, the parallel workspaces, the history tracking. The labs have spent two years racing on model quality, and now they are buying the tools that make those models usable for real development work.

That race is visible across the developer ecosystem. GitHub Copilot Fusion now picks models by effort level, GitHub Agentic Workflows lets you describe a job in Markdown and hand it to an agent, and Cursor opened its plugin ecosystem with an official spec. Every platform is moving from one agent in a chat box to teams of agents doing real work, and the winners will be the ones whose orchestration holds up. Meta just bought one of the best orchestration teams available. The price is undisclosed, but the intent is not.