The open-source home for Conductor users
Meta has acquired Conductor. CoderBaba gives you the same parallel-agent workflow from your terminal: free, open source and running any open-source model, locally or in our cloud.
step 0 · do this first
Back up your Conductor cloud workspaces
Per TechCrunch, Conductor's cloud service shuts down on November 20, 2026 and the company will then begin deleting user cloud data. Anything that only exists in a cloud workspace needs to be pushed somewhere you own. It takes about ten minutes.
- 1List every Conductor workspace that lives in the cloud, not on your laptop.
- 2In each one, commit work in progress, even if it is messy, on its own branch.
- 3Push every branch to a git remote you control (GitHub, GitLab, Origin or self-hosted).
- 4Copy any prompts, task notes or agent instructions you want to keep into the repo.
- 5Verify by cloning the remote fresh on another machine and checking out each branch.
What happened to Conductor
On November 4, 2026, TechCrunch reported that Meta has acquired Conductor, the AI coding startup that let developers spin up software projects by deploying multiple agents into parallel workspaces. The news was first reported by Axios, and Conductor announced it on its own blog. Founder and CEO Charlie Holtz wrote that the team has "big, crazy plans for the next interface" and expects to ship something new for customers in early 2027.
For existing users, three details matter most:
- Core features keep working for now. Holtz said Conductor's core features would continue to work as before until early next year.
- The cloud service does not. Conductor's cloud service shuts down on November 20, 2026, and the company will then begin deleting user cloud data.
- What comes next is a Meta product. Meta launched its own coding agent, Muse Code, in August, and Holtz said he is looking forward to working with Alexandr Wang and Nat Friedman of Meta Superintelligence Labs. Whatever ships next will follow Meta's roadmap, not necessarily yours.
None of this is a criticism of Conductor or Meta. Acquisitions are how many great startups end. It is a reminder, though, that a closed tool is only as durable as the company behind it.
Why Conductor users loved it
Conductor got a key idea right: one task, one isolated workspace, one agent. Instead of a single assistant juggling everything in one checkout, you could hand separate, compartmentalized parts of a project to separate agents, running in the cloud, locally or from your phone, and review the results side by side. The agents themselves came from major labs such as Anthropic and OpenAI. Conductor was the orchestration layer around them.
That pattern is worth keeping. The good news is that the core of it doesn't depend on any single company. It is built on git, which you already own.
Switch to CoderBaba
CoderBaba is a free, open-source AI coding agent for the command line, built around the same one-task-per-workspace idea. Install it with one command:
npm install -g coderbabaThen start tasks from your project, as many as you like:
cd my-app
coderbaba "add OAuth login"
coderbaba "fix retries in the billing webhook"Here is how your Conductor habits map over:
- Parallel workspaces. Every CoderBaba task runs in its own git worktree on its own branch. Tasks can't overwrite each other's files, and your main branch stays untouched until you merge.
- Reviewing agent work. CoderBaba shows every command and edit as it happens. Each finished task arrives as a diff you approve, reject or send back for another pass.
- Models. Conductor ran agents on hosted models from big labs. CoderBaba runs any open-source model, such as Qwen3-Coder for heavy agentic work or Devstral for a strong model that fits on one GPU. Run them on your own machine or in our free cloud.
- Cost. CoderBaba is free to use.
- Durability. CoderBaba is open source, so the tool you adopt can't be acquired out from under you, and your branches and history always live in your own git repository.
Prefer to wire it up yourself?
If you'd rather assemble your own stack, you can rebuild a Conductor-style workflow from open-source parts. Here's how.
1. Git worktrees for isolated workspaces
A git worktree is a second, third or tenth checkout of the same repository, each on its own branch, all sharing one .git history. It is exactly what you need to keep parallel agents from stepping on each other's files.
# one workspace per task, each on its own branch
git worktree add ../myapp-auth -b feat/auth
git worktree add ../myapp-billing -b feat/billing
git worktree add ../myapp-docs -b chore/docs
git worktree list # see them all
git worktree remove ../myapp-docs # clean up when mergedEach directory is a normal folder. Open it in any editor, run its tests independently, and merge its branch with a regular pull request.
2. An open-source agent in each workspace
Pick an agent you can inspect, fork and keep:
- Aider: a terminal pair programmer that edits files and commits each change to git, which makes every agent step easy to review or revert.
- OpenHands: an autonomous software-engineering agent that runs commands, edits code and works through multi-step tasks in a sandbox.
- Cline: a VS Code extension that plans, edits files and runs terminal commands with your approval at each step.
- Qwen Code: a command-line agent from the Qwen team, tuned for the Qwen3-Coder models.
Run one agent per worktree in its own terminal tab or tmux pane. That is the whole trick.
cd ../myapp-auth
aider --model ollama_chat/qwen2.5-coder:32b
cd ../myapp-billing
aider --model ollama_chat/devstral3. Models you control
Conductor ran agents powered by hosted models from big labs. You can keep doing that, since every agent above accepts API keys, but you no longer have to. Open-weight models are now good enough for a large share of everyday agent work:
- Qwen3-Coder for heavy, long-context agentic tasks
- Devstral for a fully local agent on a single 24 GB GPU or a 32 GB Mac
- Qwen2.5-Coder for fast, cheap edits and autocomplete
Serve them with Ollama on a laptop or vLLM on a GPU server, and point your agents at the OpenAI-compatible endpoint. That keeps cost down and keeps your code on your machine.
The DIY route works, but you maintain the glue yourself: creating and cleaning up worktrees, starting agents, tracking which one is doing what. CoderBaba does all of that for you.
Conductor vs. CoderBaba vs. DIY
| Conductor (as reported) | CoderBaba | DIY open-source stack | |
|---|---|---|---|
| Parallel isolated workspaces | Yes | Yes, each task in its own git worktree | Yes, worktrees you manage by hand |
| Where it runs | Cloud, local, phone | Your terminal, with models local or in our free cloud | Your machine or your servers |
| Models | Agents from major labs such as Anthropic and OpenAI | Any open-source model | Whatever each tool supports |
| Source code | Closed | Open | Open |
| Future roadmap | Set by Meta | Public, set in the open | Set by each project |
| Cloud data deadline | Cloud shuts down Nov 20, 2026 | None: your work lives in your git repo | None |
A migration plan you can finish this week
- Today: complete the backup checklist above and verify each branch on a fresh clone.
- Today, ten minutes later: run
npm install -g coderbabaand give CoderBaba one small task in a recovered branch. - This week: run two or three CoderBaba tasks in parallel on real work. Try a couple of models from our model guide to see which fits your codebase.
- Ongoing: keep everything in git. Your branches, history and workflow stay yours, whatever happens to any tool.
Moving off a tool you like is never fun, especially on a deadline. The upside is that this is the last time a vendor decision has to force a migration on you.
Conductor migration FAQ
When does Conductor's cloud service shut down?
According to TechCrunch's report of Meta's acquisition, Conductor's cloud service shuts down on November 20, 2026, after which the company will begin deleting user cloud data. Core features are expected to keep working until early next year.
Does Conductor stop working completely?
Not immediately. Conductor's founder said core features would continue to work as before until early next year. The cloud service is the exception and closes on November 20.
Is CoderBaba available today?
Yes. Install it with npm install -g coderbaba and run it from any project. It's free and open source.
Which models can I use after switching from Conductor?
CoderBaba runs any open-source model, locally on your own hardware or in our free cloud. Agentic open models such as Qwen3-Coder and Devstral handle a large share of everyday agent work.
Does CoderBaba have parallel workspaces like Conductor?
Yes. Every CoderBaba task runs in its own git worktree, a separate checkout of your repository on its own branch, so parallel tasks never overwrite each other's files.
Why choose an open-source agent after an acquisition?
Open-source code cannot be withdrawn. If a company changes direction, the tool, your configuration and your workflow keep working, and anyone can fork and maintain them.
Switch to CoderBaba in one command
Parallel workspaces, any open-source model and every diff reviewed by you. Free and open source.
$ npm install -g coderbaba