Meta Acquires Conductor, the Parallel AI Coding Startup, as Its Cloud Service Nears Shutdown

Meta has acquired Conductor, an AI coding startup founded in 2024, and confirmed the deal to TechCrunch on November 4, 2026. Founder and CEO Charlie Holtz announced that Conductor is joining Meta.

The product lets users start software projects by deploying multiple AI agents into parallel workspaces. An agent is a program built around a large language model that carries out a multi-step task with limited supervision. Workspace isolation means each agent works in its own separate copy of the code, so its changes do not collide with the others. Conductor's agentic teams can run in the cloud, locally, or be deployed from a user's phone. The team structure is built for software engineers who want separate, self-contained parts of a project worked on together, with each agent handling its own slice before the results are brought together.

The agents themselves are not developed by Conductor. According to TechCrunch, they come from major model labs such as Anthropic and OpenAI. Conductor's contribution is the orchestration layer: the workspace isolation, the coordination of parallel agents, and the interface through which engineers direct them.

The transition has a fixed timetable for existing customers. Conductor's cloud service will shut down on November 20, and the company will begin deleting user cloud data after that. Users who depend on the hosted deployment path have a little under three weeks from the date of the announcement to move work or retrieve material they need. The announcement did not change the local and phone-based deployment options, though the cloud path is the one with a hard end date.

Meta's AI leadership is led by Alexandr Wang, who serves as chief AI officer. Nat Friedman is head of product at Meta's Superintelligence Labs. Both are part of the company's broader AI organization, which is the context in which a coding-agent startup of this kind now sits.

In my view, the more useful question for engineering teams is not whether parallel agents work, but who owns the coordination layer once it becomes part of a large platform. Conductor's value sat in how it routed work across models it did not build. That makes the relationships with model providers, and the terms under which they continue, a central part of what the product is. Teams that adopted the tool for its workflow should watch how the integration develops, and whether the local and phone-based paths keep the character they had as independent offerings.

The broader context here is a shift in where agentic development tooling is being assembled. Parallel agent workspaces, which let one engineer direct several bounded tasks at once, have moved quickly from experiment to product. Folding that layer into a major platform could put the workflow in front of a much larger base of developers, and that reach is the part of the story with the most room to matter. The near-term consequence is narrower: a cloud service closing on November 20, followed by a data deletion window.