Moving from Conductor? Keep your workflow.
Everything you loved about running several coding agents at once, plus the engineering system that makes their work production-ready. Your worktrees, scripts and notes come with you.
- Nov 4No new Conductor sign-ups or downloads.
- Nov 13Conductor Cloud becomes read-only: you can read and export old chats, but not send new messages.
- Nov 20Conductor Cloud shuts down and deletion of Cloud data begins. Export before this date.
- Early 2027The local Mac app begins winding down.
From Conductor's announcement. To keep Cloud work, right-click a Cloud workspace and choose "Transfer to local", or use Settings, Account, Export data, before Nov 20.
Move in three steps
Your repository works as it is: Palgu reads your Conductor settings directly. The import adds your workspaces as tasks.
- Install Palgu
curl -fsSL https://api.palgu.ai/install.sh | bash
Or download Palgu Studio for Mac or Windows.
- Import your Conductor workspaces
palgu import conductor --dry-run palgu import conductor --apply
The dry run lists every change first. In Palgu Studio, the same import is in onboarding and Settings. Running it again changes nothing, and it never deletes or changes Conductor's data or your uncommitted work.
- Done
Each workspace is a Palgu task on the same branch, with its diff intact, its .context notes and MCP servers. Secrets go to the macOS Keychain or Windows Credential Manager, never into files.
Side by side
What you use in Conductor, from Conductor's own docs, and where it is in Palgu.
| In Conductor | In Palgu |
|---|---|
| One workspace per git worktree, with its own branch (source) | Each task or agent gets its own git worktree and branch. Your existing Conductor worktrees are adopted in place, on the same branch. |
| Claude Code, Codex, Cursor and OpenCode in parallel (source) | Claude Code and Codex run on your own login. palgu runners lists the agent CLIs this machine can use, OpenCode included. Cursor: coming once its terms check passes. |
| .conductor/settings.toml scripts: setup, run actions, archive, run_mode (source) | Read as it is, with the older conductor.json. No conversion step. |
| CONDUCTOR_WORKSPACE_PATH, CONDUCTOR_ROOT_PATH, CONDUCTOR_WORKSPACE_NAME, ten ports from CONDUCTOR_PORT (source) | The same variables, plus PALGU_ twins, with a block of ten ports per worktree. |
| Files to copy, such as .env.local (source) | Files to copy and .worktreeinclude are both honoured when a worktree is set up. |
| Notes and handoffs in the workspace .context folder (source) | Imported into the task memory, so the next agent starts from them. |
| Diff viewer with line comments, then Create PR (source) | Studio review: the diff, comments back to the agent, Create PR and merge. |
| Checks: git status, PR, CI, review threads, todos (source) | The same, plus tests, screenshots and an independent council review before work counts as done. |
| Plan Mode and Fast Mode (source) | Plan checkpoints in every task, and palgu --fast for a quick run that keeps them. |
| Skills for Codex, Claude Code and OpenCode (source) | CLAUDE.md, AGENTS.md, skills and .mcp.json are read as they are; MCP servers from Conductor are imported. |
| Cloud workspaces: each agent in its own cloud sandbox (source) | Not yet: Palgu has no hosted agent compute. Transfer the workspace to local and the agents run on your own Mac or PC, while tasks, proof and code knowledge live in your organization's hosted workspace. |
And what Palgu adds:
- Independent verification before anything counts as done
- A shared code graph and product memory for every agent
- Multi-model routing with budgets, when you want it
- Policy, a security heartbeat and an audit trail
- Palgu Studio for Mac and Windows, and the palgu CLI
Studio, the palgu CLI, parallel agents, verification, review and tasks are free when your agents run on your own Claude Code, Codex or OpenCode plan (Cursor: coming once its terms check passes), for accounts created from this page. Each agent runs through its own CLI, logged in on your machine, as it did in Conductor. Palgu never holds, proxies or replays your subscription token.
Routed models through Palgu (Mixture of Models and model groups) are optional and paid: provider cost plus a 10% gateway fee, with no seat charge. See pricing.
A day on Palgu
The work you did in Conductor, step by step, in Studio or in the terminal.
In Studio, start three tasks; each gets its own worktree and branch. From the terminal, start them in the background and watch them in one view.
palgu agents run implementer "add the search endpoint" --worktree --detach palgu agents run implementer "fix the login redirect" --worktree --detach palgu agents run implementer "write the export docs" --worktree --detach palgu agents
Open a task in Studio to see its diff and comment on a line to send it back to the agent. palgu review sends the branch to the review council, which checks each finding against your code before it reports it.
palgu review main
Create the pull request and merge from Studio's review. palgu ship first tells you whether the commit is clear to ship.
palgu ship
Pick up your last session, or a paused task by its id. A Claude Code or Codex session from your Conductor workspace continues on the same CLI.
palgu --continue palgu resume <id> palgu --continue-claude
Palgu writes the test first and runs it. When a test fails, the agent sees the failure in the same task and fixes the code; the task does not move on until the tests pass.
palgu -p "the checkout total test fails: fix it"
Before work counts as done, Palgu checks it independently: tests, screenshots for anything with a screen, a council review and evidence tied to the commit. You see what was checked.
palgu agents run verifier "check the search endpoint does what the task says" palgu check quality
More in the docs.