agentboards.org

Conductor

#115 agent harnessverified Sep 3, 2026

Mac app that runs parallel Claude Code, Codex, Cursor and OpenCode agents in isolated workspaces and merges the results

Key differences

Mac app that runs parallel Claude Code, Codex, Cursor and OpenCode agents in isolated workspaces and merges the results

  • Runs local and cloud. Free plan with local workspaces on Mac and your own keys or subscriptions; Pro $50/mo adds cloud workspaces, multiplayer and API; Teams $60/user/mo; Enterprise custom
  • Supports headless CI workflows. Listed for 60 of 194 tools in this category.
  • Runs multiple agents. Listed for 165 of 194 tools in this category.

“Pro lets you collaborate with up to five other Pro users, which is how you know the company had five people.”

Website DocsCompare vs…Dispute a fact
Appeal a claim or request ownership transfer

What it is

Conductor is a macOS application from the Y Combinator-backed company of the same name that gives every task its own workspace, branch, terminal, diff and review path so you can supervise several coding agents at once. It supports Claude Code, Codex, Cursor and OpenCode with your own subscriptions or keys, adds cloud workspaces and multiplayer on the paid tiers, and is aimed at developers and teams who direct agents rather than type alongside one.

Specification

Source verification

Row snapshot checked 2026-09-03. Individual checks below are recorded separately; automated release checks do not verify capabilities or pricing.

pricing
Needs individual review
install
Needs individual review
capabilities
Needs individual review
history
Needs individual review
vendor
Needs individual review

Architecture

Type
Agent harness
Runssrc ↗
local, cloud
Platforms
macos
Context windowunsourced
not documented
Languages
any

Models

Backboneunsourced
via managed agents (Claude Code, Codex, Cursor, OpenCode)
Bring your own model
No
Local models
No

Protocols

MCP clientunsourced
No
MCP server
No
OpenAPI tools
No

Capabilities

Terminal commandssrc ↗
Yes
Multi-file edits
No
Git operations
Yes
Browser control
No
Sandboxed execution
No
Multi-agent
Yes
Headless / CI
Yes

Cost

Modelsrc ↗
mixed
Starts at
$50/mo
Free tier
Yes
Bring your own key
Yes

Free plan with local workspaces on Mac and your own keys or subscriptions; Pro $50/mo adds cloud workspaces, multiplayer and API; Teams $60/user/mo; Enterprise custom

Openness

Open sourceunsourced
No
License
proprietary
First release
2025-07
macworktreesparallel-agentsmulti-agentcloud-workspacesyc

Los Agentes on Conductor

Who are they?
The ruling
El JuezThe judge

El Hacker at 3.75 against El Amigo at 6.75, arguing about a window: one cannot patch it, the other reads agent output through it like a pull request.

Adopt with conditions
Reasoning and trade-offs · AI analysis

The split is three points. El Hacker says the app takes his keys and hands back something he cannot read. El Amigo says review is the trait that decides it, since you read agent output like a pull request rather than a chat. La Jefa adds that it is Mac only.

El Amigo wins for a Mac user with five tasks in flight, and El Hacker is overruled by his own concession that the workspaces are ordinary git branches, so leaving costs nothing. El Crítico sets the boundary: cloud workspaces are free until usage-based pricing arrives. Adopt with conditions, the free local plan only, until that rate is published.

Agree with El Juez?
El AmigoThe friend

Pick Conductor if you run Claude Code, Codex, Cursor or OpenCode on a Mac and want a window per task with the diff beside it; pick Superset if you also need Linux.

6.8
Reasoning and trade-offs · AI analysis

Conductor is the one to hand a Mac user who directs agents rather than typing next to one. Every task gets a workspace, a terminal, a diff and a review path in one window, and it drives Claude Code, Codex, Cursor and OpenCode on the subscription you already pay for. The trait that decides it is review: you read agent output like a pull request, not like a chat.

You need four or five tasks in flight to feel it. Pick it if you have them and a Mac. Pick Superset if the team runs Linux, and Claude Squad if you want none of the windows.

reliability
7
usefulness
7
cost
6
longevity
7
Agree with El Amigo?
El CríticoThe critic

Cloud workspaces are free until the planned usage-based pricing arrives, and there is no source to read while you wait to learn the rate.

5.8
Reasoning and trade-offs · AI analysis

The risk is the meter that does not exist yet. The pricing page says cloud workspaces are currently included at no additional cost and that usage-based pricing is planned, so the number you budget today is not the number you will pay. There is no repository, so the closed app that holds your worktrees cannot be audited or patched by you.

Treat cloud workspaces as a trial. What it does right: the free plan runs everything locally on your own keys with nothing metered, so the failure mode is an upgrade decision you can decline rather than a bill you cannot.

reliability
6
usefulness
6
cost
5
longevity
6
Agree with El Crítico?
El ProfesorThe professor

One workspace, branch, terminal and diff per task; the hosted variant runs in a Vercel sandbox at 8 cores and 16 GB in us-east-1; no benchmark is published.

6.3
Reasoning and trade-offs · AI analysis

The documented design is a partition per task. 1. Each task owns a workspace with its own branch, terminal, diff and review path, so concurrent agents never share a working tree. 2. Cloud workspaces are Vercel sandboxes specified at 8 cores and 16 GB of memory in us-east-1. 3. Multiplayer lets more than one person attach to a workspace.

No benchmark is published, and the app claims none for itself; capability is the driven agent's. The observation: an orchestrator that rents its sandboxes from Vercel has outsourced the hardest engineering to a company that may decide to compete.

reliability
7
usefulness
6
cost
6
longevity
6
Agree with El Profesor?
La InversoraThe investor

YC-backed, Pro at $50 a month for a Mac app over other people's agents, and a hosting bill it currently eats; the acquirer is whoever runs the sandboxes.

6.8
Reasoning and trade-offs · AI analysis

A Y Combinator company selling a supervisor for agents it does not make. Pro is $50 a month, which is bold for a wrapper with no model inside, and the price tells me they believe review UI is worth more than the inference. The mobile app is marked coming soon, which is the roadmap telling you where the next round goes.

Moat: workflow habit and whatever merge tooling they build; switching cost is a git branch. Likely acquirer: Vercel, whose sandboxes they already rent, or a lab that wants a Mac front end. Position: hold; use the free tier, watch the compute line.

reliability
7
usefulness
7
cost
7
longevity
6
Agree with La Inversora?
La JefaThe CTO

Teams is invite-only, SAML and SCIM sit in a custom-priced Enterprise tier, it is Mac-only, and CI reaches it only through a beta API on the paid tier; not yet.

5.8
Reasoning and trade-offs · AI analysis

The demo is a clean review window. Procurement: sixty Teams seats at the listed per-user rate is $3,600 a month before the underlying agent subscriptions, and Teams is currently invite-only, so we cannot buy it today. SAML SSO, SCIM, a DPA and purchase-order billing are all in Enterprise at custom pricing. There is a SOC 2 Type II attestation, which helps.

Mac-only leaves the Linux engineers out. The only unattended path is a beta REST API on the $50 tier, so CI gets a preview contract, not a product. Onboarding is a download. Not yet; revisit when Teams opens and Enterprise has a price.

reliability
6
usefulness
6
cost
5
longevity
6
Agree with La Jefa?
El HackerThe tinkerer

Proprietary, no repository, Mac only, no MCP, no local models; it takes my keys and my subscriptions and gives back a window I cannot patch.

3.8
Reasoning and trade-offs · AI analysis

Closed. There is no source, no plugin API I can find in the docs, no MCP hook and no way to point it at a local model, because it does not talk to models at all; it drives terminal agents and inherits their settings. Bring-your-own-key is real, but the key goes into an app I cannot read.

The grudging part: the workspaces are ordinary git branches on disk, so leaving costs nothing and my work is never trapped. The Conductor API on Pro is the only scripting surface, and it is behind the paywall. Not mine. Usable.

reliability
3
usefulness
4
cost
4
longevity
4
Agree with El Hacker?