Meta didn’t buy a model this time — it bought Conductor, in an acquisition disclosed November 4, 2026: the startup whose entire product is running Claude Code, OpenAI’s Codex, Cursor, and OpenCode in parallel, isolated workspaces, then letting a developer review and merge whatever comes back. Axios broke the deal exclusively, and Conductor co-founder and CEO Charlie Holtz confirmed it himself on X: “Conductor is joining Meta.” Terms weren’t disclosed.

The shutdown notice was disclosed, though, right down to the day. Conductor Cloud — the hosted version of the product, the one that made running four different vendors’ coding agents at once something you could do without provisioning your own infrastructure — goes dark on November 20. That’s 16 days after the acquisition was announced, for a product with more than 100,000 users.

Quick Summary: What Happened

Detail	Info
Date announced	November 4, 2026, per Axios
What Meta bought	Conductor — a multi-vendor orchestration layer for Claude Code, Codex, Cursor, and OpenCode agents
Users affected	More than 100,000
Conductor Cloud shutdown	November 20, 2026 — 16 days after the deal was announced
Pre-acquisition funding	$22M Series A, March 2026, led by Spark Capital and Matrix
New home for the team	Meta Superintelligence Labs, under Nat Friedman and Alexandr Wang
Confirmed by	Charlie Holtz on X and Conductor’s own announcement

Bottom line: Meta bought the coordination layer sitting on top of its rivals’ coding agents, then shut down the one product that made that neutrality useful to 100,000 people — with barely two weeks’ runway to get out.

What Actually Happened in the Meta-Conductor Acquisition

Conductor doesn’t build models. That’s the whole pitch. Per Conductor’s own product documentation, the software spins up Claude Code, Codex, Cursor, and OpenCode agents in their own sandboxed workspace — isolated copy of the repo, own branch, own terminal, own diff — so a developer can hand the same task to two or three agents at once and compare what each one does with it, or split a project into pieces and let different agents work in parallel without stepping on each other’s files. Users bring their own subscriptions and API keys to whichever vendor they want. Conductor coordinates. It doesn’t compete on model quality, because it has no model to defend.

That positioning is exactly why Meta wanted it, and exactly why the acquisition is uncomfortable to look at directly. Founded in 2024 and launched publicly in the summer of 2025, Conductor had crossed 100,000 users and raised a $22 million Series A in March 2026 from Spark Capital and Matrix before Meta came calling. TechCrunch’s reporting on the deal notes Meta confirmed the acquisition but declined to comment further — standard practice, but a thin statement given what the company is actually walking away with: a working, proven interface for running Anthropic’s and OpenAI’s flagship coding agents side by side.

Holtz framed the move as expansion, not an ending, in his announcement on Conductor’s site and on X: the team is joining Meta Superintelligence Labs to work directly with Nat Friedman and Alexandr Wang, and he’s promising “something new” — “big, crazy plans for the next interface” — by early 2027. What he didn’t promise is that the next interface keeps treating Claude Code, Codex, and Cursor as equals. Meta now owns the team that built a vendor-neutral coding layer. Meta is not vendor-neutral.

What Happens to Conductor Cloud Users Now?

For the 100,000-plus people using the hosted product, the wind-down runs on a tight, published clock:

1. November 4 — new sign-ups stopped the day the acquisition was announced, and existing free users can no longer upgrade to a paid plan.
2. November 13 — Conductor Cloud goes read-only. Existing workspaces and chat history stay viewable and exportable, but you can’t send a new prompt to an agent after this date.
3. November 20 — Conductor Cloud shuts down fully, and the company begins deleting cloud user data. This is the hard deadline.
4. Active subscribers get refunded for the unused portion of their plan rather than being billed through a service that no longer exists.
5. The local Mac app keeps working as normal for now, with its own wind-down not starting until early 2027 — a materially longer runway than the cloud product got.

If you’re one of the 100,000, the move is to export anything in Conductor Cloud before November 13, and decide this week: shift work to the local app, which has months more runway, or move off the platform entirely.

Why This Matters

Meta’s recent acquisitions have followed a pattern: buy the team, fold it into Meta Superintelligence Labs, point it at a Meta-controlled product. That pattern makes sense for a robotics startup or a research lab. It’s a stranger fit for a company whose entire value proposition was refusing to pick a side between Anthropic, OpenAI, and the rest.

This is the same shape of problem we flagged when SpaceX bought Cursor back in June: a tool’s whole appeal was that it didn’t care which model you used, and then a company with a model to sell bought the tool. SpaceX didn’t shut Cursor down — it’s a much bigger, standalone business with $4 billion in ARR, not a feature Meta is folding into its own stack — but the incentive is identical. Once the acquirer has a horse in the race, “doesn’t matter which agent you pick” stops being a feature and starts being a problem to solve. Meta shutting down Conductor Cloud within 16 days, rather than running it as an independent product the way SpaceX kept Cursor running, suggests Meta isn’t interested in operating a neutral marketplace for competitors’ agents at all. It wants the orchestration idea, not the business of hosting Claude Code and Codex for a six-figure user base.

That’s a meaningful signal for anyone building workflows around Claude Code or comparing it against Cursor and GitHub Copilot. The orchestration layer — the part of the stack that decides which agent runs which task — is turning out to be just as contested as the agents themselves, and just as likely to get absorbed into a single vendor’s roadmap the moment it gets big enough to notice.

What Are Your Options Now

If you’re migrating off Conductor Cloud before the November 20 shutdown:

• Export everything by November 13. Once Cloud goes read-only, you can still read and export, but the clock to actually move data runs out fast after that.
• Move active projects to Conductor’s local app if you want to keep the same multi-agent workflow without an immediate deadline — it’s not getting wound down until early 2027.
• Evaluate alternatives built on the same multi-agent premise. Competing orchestration tools have already started courting displaced Conductor users with migration guides, a sign the gap Meta is leaving isn’t going unnoticed.
• If your workflow depended specifically on running Claude Code and Codex side by side, check whether your usage falls under each vendor’s own multi-agent tooling before assuming you need a third-party orchestration layer at all — both Anthropic and OpenAI have been building more of this natively over the course of 2026.

The Bigger Picture

Conductor isn’t the only developer-tools deal that closed the week of November 4. Axios also reported that OpenAI told the outlet, the same day, that it plans to acquire Pierre Computer, a developer-tools startup whose Git-based code storage technology OpenAI wants to fold into Codex. That follows OpenAI’s acquisition of Ona (formerly Gitpod) back in June, which gave Codex persistent cloud sandboxes instead of ephemeral, session-bound ones. Two frontier labs, same week, same instinct: whoever owns the infrastructure coding agents actually run on owns a harder-to-dislodge piece of the developer’s workflow than whoever owns the model underneath.

Meta’s own stake in that fight isn’t theoretical. Meta’s own Muse personal AI agent, which launched September 8, 2026, reportedly reached 10 million U.S. downloads within 40 days. Conductor’s team — people who’ve already solved the UX problem of running several agents in parallel without the whole thing turning into chaos — is a shortcut to Meta’s next move in that product, not a side project Meta plans to leave running independently.

For the rest of the industry, this is the clearest evidence yet that developer tooling, not just model capability, is where Claude Code, Codex, Cursor, and Meta’s own agents are actually going to compete in 2027. The orchestration layer briefly looked like neutral ground. It’s not going to stay that way.

Our Take

I don’t think Meta bought Conductor because it needed another product to sell. I think it bought Conductor because the team had already figured out the hard UX problem of making multi-agent coding feel coherent instead of chaotic, and that’s a genuinely harder problem than people assume — ask anyone who’s tried to run two coding agents on the same repo without them clobbering each other’s work.

Where I’d push back on the framing in Holtz’s own announcement: calling this an expansion elides what’s actually happening to the 100,000 people who built workflows around Conductor Cloud specifically because it was vendor-neutral. Sixteen days is not generous notice for a hosted product with six-figure usage, even with refunds and a local-app fallback. Compare that to SpaceX, which kept Cursor running as an independent business after its acquisition closed. Meta chose to shut the comparable product down fast instead. That’s a company that wanted the team and the architecture, not the business of hosting a tool that treats Claude Code and Codex as equally good options.

The honest read is that “multi-agent orchestration” was never going to stay neutral once a frontier lab owned it. Anthropic, OpenAI, and now Meta all have the same incentive once they’re the ones running the coordination layer: make their own agent the default, not one option among four. Conductor’s users got a preview of that incentive in the fastest, least comfortable way possible — a shutdown notice instead of a roadmap update.

Frequently Asked Questions

What is Conductor, and what does Meta’s acquisition include?

Conductor is a multi-vendor orchestration platform that runs Claude Code, OpenAI’s Codex, Cursor, and OpenCode agents in parallel, isolated workspaces, letting a developer compare or split work across them. Meta acquired the company on November 4, 2026, for undisclosed terms, bringing the founding team into Meta Superintelligence Labs. Conductor builds no models of its own — it coordinates other companies’ agents.

Why is Conductor Cloud shutting down?

Meta hasn’t given a public reason beyond folding the team into Meta Superintelligence Labs to build a new product. Conductor’s own shutdown notice frames it as a transition rather than an abandonment, with CEO Charlie Holtz promising a new interface by early 2027. The practical effect, regardless of framing, is that the hosted product 100,000-plus users relied on ends November 20, 2026.

When exactly does Conductor Cloud stop working?

New sign-ups stopped November 4, 2026. The service goes read-only on November 13. It shuts down completely, with cloud user data deleted, on November 20 — 16 days after the acquisition was announced. Conductor’s local Mac app keeps working normally until it begins its own wind-down in early 2027.

Will I get a refund if I was paying for Conductor Cloud?

Yes. Per Conductor’s own shutdown announcement, active subscribers receive refunds for the unused portion of their plan rather than being billed through a service that’s shutting down.

Who funded Conductor before the acquisition?

Conductor raised a $22 million Series A in March 2026 led by Spark Capital and Matrix, following earlier backing through Y Combinator. The round came about eight months before Meta’s acquisition.

Is this part of a bigger trend of AI labs buying developer-tools startups?

Yes. The same week Meta acquired Conductor, OpenAI told Axios it plans to acquire developer-tools startup Pierre Computer to add code storage to Codex — following OpenAI’s June 2026 acquisition of Ona (formerly Gitpod) for persistent cloud sandboxes. Frontier labs are increasingly competing on the infrastructure coding agents run on, not just on model quality.

Should I keep using Claude Code, Codex, or Cursor directly instead of a third-party orchestration layer?

That depends on whether you need to run multiple agents on the same task, which is Conductor’s specific value. If you were using Conductor mainly as a convenience wrapper around a single vendor’s tool, going direct removes a dependency on a product mid-transition. If multi-agent comparison was the point, you’ll need to evaluate what replaces that workflow before November 20.

---

Last updated: November 6, 2026. Sources: Axios — Meta acquires AI coding startup Conductor · TechCrunch · SiliconANGLE · Charlie Holtz on X · Conductor’s own announcement · Conductor product documentation · OpenAI — OpenAI to acquire Ona · DevOps.com — OpenAI Acquires Pierre Computer.

Related reading: SpaceX Buys Cursor for $60B: What Devs Must Know · Cursor vs Claude Code vs GitHub Copilot 2026 · Claude Code Routines: What Enterprises Should Know · Meta’s ARI Acquisition: Physical AI Enters Enterprise · DeepSeek’s Super-Agent Takes Aim at Claude Code