brindle vs Conductor
Both brindle and Conductor help run coding agents across separate Git worktrees. They differ in where the workflow lives: brindle is a terminal-based supervisor that delegates and verifies work, while Conductor is a Mac app for managing workspaces, agent sessions, and review.
At a glance
| Area | brindle | Conductor |
|---|---|---|
| Primary form | Source-available terminal application with a tmux chat and status sidebar. | Mac application organizing repositories into workspaces, chats, terminals, diffs, and review state. |
| Parallel work | A supervisor can split goals into tasks and assign workers to separate branches and worktrees. | Create separate workspaces for independent tasks, each with its own branch and working tree; multiple agents can also share one workspace. |
| Agent options | Claude Code, Codex, Google Antigravity, and open-weight models through brindle's native agent loop. | Claude Code, Codex, Cursor, and OpenCode in workspaces. |
| Progress and completion | Milestones have check commands. A milestone is verified when its check passes; merges require passing checks and, under autopilot (on by default) or with review turned on, a reviewer's approval of the exact commit. | Run checks and review diffs in the workspace workflow. Checks can track git status, CI, deployments, comments, and todos. |
| Review and landing work | Workers report results to the supervisor, which can request a reviewer and merge a branch after its gates pass. | Review changes in the Diff Viewer, use the Checks tab, then create a pull request, merge, and archive a workspace. |
| Local environment | Runs in a repository from the terminal; workers use their own worktrees, with optional per-worktree service containers on a paid tier. | Runs on Mac. Workspaces use Git worktrees; setup and run scripts prepare and operate each workspace separately. |
| Paid capabilities | Optional capabilities include per-worktree services, hosted task-to-profile learning, sessions across several repos, CI automation, team policy, audit records, and air-gap operation. | Not compared here; see Conductor's site for its current plans. |
How the workflows differ
brindle starts with a supervisor conversation in the repository. You describe a goal, and the supervisor can divide it into milestones and tasks. Workers run in separate Git worktrees, allowing independent changes to proceed at once. The supervisor tracks worker status and milestone checks in its sidebar. A milestone does not count as done because an agent says it is done: its check command must exit successfully. Before a branch merges, the configured checks must pass, and under autopilot (on by default) or with review turned on, a reviewer must approve that exact commit. See the brindle overview for its full workflow and supported setup.
Conductor starts with workspaces in a Mac app. A workspace groups a branch and working tree with the agent chat, terminal, application processes, diff, and review state. Its docs recommend one workspace per independent task when each change should have its own review path. If agents need the same branch and current files, they can instead work in the same workspace. The app supports Claude Code, Codex, Cursor, and OpenCode. Its documented workflow covers setting up and running a workspace, reviewing the diff, tracking checks, creating a pull request, merging, and archiving the workspace.
Both products use Git worktrees to give parallel tasks their own files and branches. Conductor puts the workspace and its review tools in a desktop interface. brindle adds a goal-driven supervisor, task delegation, persistent progress tracking, and explicit completion and merge gates. That distinction matters most when deciding whether a person or an orchestrator should coordinate the work and decide when it is complete.
Coordination and verification
In brindle, workers can report results and communicate with the supervisor through brindle's agent tools. The supervisor can assign dependent tasks, wait for workers, request review, and check milestone progress. Checks can run again after later merges, so a previously passing milestone can be rechecked against the updated code. The gated merge process makes review and repository checks part of the workflow rather than only a manual handoff.
Conductor's docs emphasize workspace-by-workspace operation. A user starts agents in a workspace, runs commands there, inspects changes in the Diff Viewer, and can send comments back to an agent. The Checks tab brings together repository and pull request signals. Its workflow gives the user direct control over which workspace to review, merge, or archive. The published material describes task and workspace management, but does not describe brindle-style milestone checks that determine when a multi-step goal is complete.
For teams that want more automation around the terminal workflow, brindle has optional paid capabilities. Per-worktree services give each worktree its own service containers; hosted learning helps choose a profile for a task; brindle CI runs autopilot through milestone checks and creates a pull request; team policy can set provider, model, worker-cap, and human-review rules; and audit features record agent actions. Air-gap operation supports running against models on the local machine or private network. These are product capabilities, not needed for the local workflow. Details are on the Enterprise page.
Where Conductor fits
Conductor is a good fit if you want a Mac app to keep each task's chat, branch, working files, terminal, running app, and review surface together. Its docs give practical setup guidance for projects that need dependencies, local configuration, or separate app processes in each workspace. It also suits work where a person wants to choose the agent, inspect each diff, follow checks, and decide when to merge or archive. The same workspace can host agents that need shared branch state, while separate workspaces suit independent pull requests.
Its workspace approach also makes sense when the main need is to fan out a set of independent changes and handle them one at a time. The documentation cautions that separate workspaces still run with the local user's permissions and that workspace isolation is not a security boundary. Conductor describes itself as a Mac app, so it is aimed at people who want that desktop workflow.
When to choose each
- Choose brindle when you want a terminal-centered supervisor to break a larger goal into work, delegate to workers, track milestone checks, and gate merges on review and passing checks. It is also a fit when you want to use local or open-weight models in the same orchestration workflow.
- Choose Conductor when you want a Mac workspace app that organizes agent sessions and their worktrees, terminals, app processes, diffs, and pull request review in one place. It fits a workflow where you decide how to split tasks and personally manage review and landing.
- Consider either when the main requirement is parallel agents working on isolated branches. Compare the orchestration and review workflow you want, along with whether you prefer the terminal or a desktop app.
Sources
Conductor facts above are drawn from its documentation, including its pages on parallel agents, Git worktrees, and workflow. brindle facts are drawn from the brindle overview and Enterprise page.
curl -fsSL pawdelta.com/brindle/install | sh