Compare·Parallel agent orchestrator
Continuum
Multi-agent workbench
VS
Conductor logo
Conductor
Parallel agent orchestrator

Continuum vs Conductor

Both spin agent fleets in worktrees. One now sells the cloud around the board; the other stays free and puts the fleet in your pocket.

Conductor (Melty Labs) is a focused macOS orchestrator: Claude Code, Codex, Cursor, and OpenCode running in parallel isolated workspaces, each with its own branch, files, terminal, diff, and review path. As of August 2026 it also sells the layer around that board: a free local tier, Pro at $50/mo for Conductor Cloud sandboxes plus multiplayer and an API, Teams at $60/mo per user, and a quoted Enterprise tier. Continuum is the free multi-provider workbench around the same class of agents, on Mac, web, iPhone, and Watch, with live 5h and weekly quota gauges, multi-account subscription isolation, and a local LiteLLM-priced ledger that shows spend by repo. Conductor's board is excellent when the whole job is babysitting dense worktrees on one Mac. Continuum's bet is the same isolation model plus clients in your pocket, multi-account Max stacking, a spawn grid, and cost you can read before the weekly window collapses. Background reading: parallel Claude Code sessions.

Download for Mac All comparisons
Workflow Monday path Matrix Decide Money FAQ

Updated 2026-08-03 · Mac stable · Win/Linux desktop beta

Choose Continuum when

Pick Continuum when you want mobile plan approve that ships today, sessions across Claude, Codex, Cursor, Grok, Gemini, and OpenCode, live 5h and weekly gauges, multi-account subscription isolation, and spend by repo, on an app that costs nothing.

Choose Conductor when

Pick Conductor when the Mac board is the product: dense parallel workspaces, Conductor Cloud sandboxes that keep working after you shut the laptop, and multiplayer with shared chats, presence, and reassignment for a team on one plan.

Snapshot direct alternative
Dimension Continuum Conductor
Worktree isolation Yes · per-session worktree mode Yes · core product
Board / fleet UX Sessions sidebar + spawn grid Parallel workspace board
Multi-provider Claude PTY · Codex ACP · Cursor · Grok · Gemini · OpenCode… Claude Code · Codex · Cursor · OpenCode
Mobile Native iPhone + Watch · QR/Tailscale pair Mac app · mobile listed as coming soon
Gauges Menu-bar 5h + weekly gauges No live quota gauges
Cost ledger Local JSONL → $ by repo/provider/day No published spend ledger
Price Free app · hosted inference optional Free tier · $50/mo Pro · $60/user Teams
Mac stable · web · iPhone · Watch · Win/Linux desktop beta · free app
01

Board at the desk vs fleet in your pocket

In Continuum

Fleet day with cost and phone

09:12

Spawn three worktree sessions: Claude Max on auth, Codex on the API path, the Cursor agent on the third. The sidebar lists all three under the same repo key.

11:40

A plan lands mid-standup. Approve it from the iPhone over the paired host; the Mac keeps the session. A Watch glance shows the turn is live.

15:20

The auth branch goes sideways. Restore the pre-prompt checkpoint, re-prompt with a tighter spec, and the worktree is clean again without touching main.

17:05

The Usage tab shows which repo burned the weekly Claude window. Pin the secondary account for the evening batch with no re-login gymnastics, and read the day in dollars, not vibes.

In Conductor

Desktop board with cloud behind it

09:12

Open Conductor on the Mac. Attach agents into isolated workspaces and treat the board like a junior engineering squad.

11:40

Stay at the desk. Mid-run plans and status live on the desktop board; the mobile app is still listed as coming soon.

15:20

Push the long job into a Conductor Cloud sandbox so it survives the laptop closing. That capability starts at the $50/mo Pro tier.

17:05

Close the day with worktrees landed. Quota truth and per-repo spend stay outside the product, so the weekly window is still a surprise.

02

Monday path

Monday with both worktree tools in play

Install, connect, first session - steps you can run the same day.

01

Download Continuum for Mac. Pair the iPhone with the QR flow over Tailscale so plan-ready can leave the desk.

02

Open the Code tab and spawn a worktree session on Claude for ticket A, then a second worktree on Codex for ticket B.

03

Keep Conductor open if you like its board. Its free tier still runs local Mac workspaces on your own subscriptions.

04

At standup, approve a plan from the iPhone. The Mac keeps the session alive and the diff waits for you.

05

Before lunch, open the Diff pane on the hottest worktree and stage only the hunks you want. See reviewing AI-generated code.

06

Evening: the Usage tab shows which repo burned the weekly window. Pin the secondary Claude account for the night batch.

07

If a run has to outlive the laptop, enroll a Linux box over Tailscale instead of buying cloud hours.

08

Optional: open a spawn grid of 4, 6, or 8 tiles when you want several agent CLIs in one pane without a board.

How Continuum actually runs the fleet

How Continuum actually runs parallel agents

  • Worktree mode creates an isolated git worktree per session; Code panes attach chat, plan, diff, PR, terminal, and artifacts.
  • Claude Code spawns over PTY; Codex, Cursor, and Grok run through the ACP harness. Same official engines, Continuum-owned store.
  • Multi-account uses CLAUDE_CONFIG_DIR and CODEX_HOME isolation, not a HOME swap, so a secondary Max never clobbers the primary Keychain.
  • Every prompt takes a git checkpoint first, so a bad turn is a restore instead of an argument with the agent.
  • The session fleet projects into the sidebar, and iPhone or Watch pair over Tailscale MagicDNS remote control.
  • Local analytics price session JSONL with a LiteLLM snapshot into $ by repo, provider, and day, beside live 5h and weekly gauges.
  • The PR pane and gh flows close the loop from plan approve to open PR without leaving the Code tab.
03

Worktree fleet matrix

Cells use product nouns on both sides. Continuum is the multi-provider workbench; they win where their loop is the product.

Capability Continuum Conductor
Parallel agents in worktreesIsolated branches without dirtying main Worktree sessions + spawn grid Core product focus
Vendor-neutral multi-providerPin a provider per session Claude PTY · Codex ACP · Cursor · Grok · OpenCode… Claude Code · Codex · Cursor · OpenCode
iPhone + Apple Watch controlApprove, steer, glance Native iPhone + Watch · QR/Tailscale pair Mobile app announced, not shipped
Usage analytics & $ costBy repo, provider, day Local JSONL → $ by repo/provider/day Not a cost product
Live quota gauges5h / weekly windows Per provider + multi-account No gauges
Web + Windows + Linux clients Mac stable · web · Win/Linux desktop beta macOS app only
Agents that outlive the laptopSleep the Mac, keep the turn Enrolled Linux/Windows hosts · Cloud runners Conductor Cloud sandboxes on Pro
Team multiplayer + presence Org accounts · shareable transcripts Shared chats, presence, reassign
SSO, admin portal, SCIM Org path Teams $60/user · Enterprise
Multi-account Claude / Codex CLAUDE_CONFIG_DIR / CODEX_HOME isolation No published account isolation
Code tab panesChat, plan, diff, PR, terminal, artifacts Chat · plan · diff · PR · terminal · artifacts Workspace, terminal, diff, review
Automation API Host daemon HTTP/WS API + CLI client Conductor API on Pro
Free tier · BYOK Free app · optional Plus/Max/Ultra hosted Free local tier · your subs and keys
04

Where the products actually diverge

01 · Surfaces

Conductor is a desk with a cloud behind it. Continuum is every screen.

Conductor optimizes the multi-agent macOS desktop, and its Pro tier lists a mobile app as coming very soon. Continuum ships the phone today: native iPhone and Watch clients, plus web and a beta Windows/Linux desktop, all opening the same session id. Leave the keyboard mid-run and Continuum still has a control surface in your pocket. More on that in mobile control.

02 · Economics

Agents are expensive. Only one product shows the bill.

Continuum prices local session JSONL with a LiteLLM snapshot and shows spend by repo, provider, and day beside live 5h and weekly gauges. Conductor's meter is its own cloud workspace hours and seats, which is a different question from "which repo burned my Claude weekly window?". If that question matters, see how the limits actually work.

03 · Providers

One fleet, many labs.

Conductor's provider list is real: Claude Code, Codex, Cursor, and OpenCode in parallel. Continuum runs those and adds Grok, Gemini, Antigravity, Z.ai, and OpenRouter, pins a provider per session, and lets two agents race the same ticket in separate worktrees so you keep the better branch. Field guide: AI coding agents compared.

04 · Team shape

Conductor sells the squad. Continuum sells the operator.

Multiplayer chats, presence, task reassignment, an admin portal, SSO, and SCIM are Conductor product, priced at $60/mo per user and up. Continuum's free app is built around one operator with many machines and many accounts, with shareable read-only transcripts when someone else needs to see a run. If you are buying seats for a squad that wants to watch each other's agents, that is Conductor's lane and we will not pretend otherwise.

05 · Worktrees

Both take isolation seriously.

If the whole job is spinning five worktrees and babysitting a junior fleet on one Mac, Conductor's board is purpose-built. Continuum also spawns isolated worktrees and multi-tile spawn grids, then adds mobile, multi-account, analytics, and multi-provider depth around them.

05

Coexistence

Stack recipe

How people run both.

Stack recipe: keep Conductor as the specialty desktop board for dense multi-worktree babysitting, and pair Continuum for iPhone plan approve, multi-account gauges, spawn grid, and cross-provider Code. Conductor's free tier still runs local Mac workspaces on your own subscriptions, so the pairing costs nothing extra until you want its cloud hours. Neither product asks you to uninstall the other, because git worktrees are a shared substrate, not a lock-in.

Download Continuum Providers
06

Pick for the job, not the category

Continuum

Choose Continuum when ops leave the desk

  • Run the fleet from Mac, web, and iPhone today, not from a mobile app that is still coming soon.
  • See live 5h and weekly gauges with multi-account Claude and Codex isolation.
  • Read spend by repo, provider, and day from a local ledger you own.
  • Keep Claude, Codex, Cursor, Grok, and peers in one Code tab with full panes.
  • Stay on a free app with the subscriptions you already pay for.
  • Prefer your own enrolled Linux box over metered cloud workspace hours.
Conductor

Choose Conductor when the board is the product

  • Live only on macOS and want a purpose-built parallel-workspace cockpit.
  • Want cloud sandboxes that keep the agent running once the laptop shuts.
  • Need multiplayer: shared chats, presence, and reassigning a task to a teammate.
  • Are buying seats with an admin portal, SSO, and SCIM rather than instrumenting one operator.
  • Do not need quota gauges or a per-repo cost ledger.
  • Want the lightest product chrome around worktree isolation.
07

What you pay

Continuum

App + your labs

$0 for the app on Mac, web, iPhone, and Watch. Bring your own keys or the Claude and ChatGPT plans you already pay for, and Continuum takes no cut. Optional hosted inference: Plus $25/mo, Max 100 $100/mo, Max 200 $200/mo, Ultra $500/mo.

Conductor

Their bill

Conductor keeps a free Mac tier: local workspaces, parallel agents, your own subscriptions and keys. As of August 2026 Pro is $50/mo and adds Conductor Cloud with a pool of cloud workspace hours, multiplayer, the Conductor API, and a mobile app listed as coming soon. Teams is $60/mo per user for the admin portal, centralized billing, and SSO. Enterprise is quoted and adds a DPA, purchase-order billing, and SCIM.

How to compare

Total cost of work

Compare what each meter counts. Continuum meters nothing on a session you run under your own login; Conductor's paid tiers meter cloud workspace hours and seats. Neither one changes what Anthropic or OpenAI bill you.

Conductor plans checked 3 August 2026 against their pricing page. Vendors reprice often - if you spot a stale figure, tell us and we will fix it. Continuum never marks up a session it runs under your own login.

Continuum ladder: Free app · Plus $25/mo · Max 100 · Max 200 · Ultra - full pricing. Also analytics, multi-account, devices.

08

Questions

Continuum vs Conductor.

Deep dives: docs, providers, sessions.

There is still a free Mac tier: parallel agents in local workspaces on your own subscriptions and keys. As of August 2026 the paid ladder above it is Pro at $50/mo for Conductor Cloud, multiplayer, and the API, Teams at $60/mo per user for the admin portal and SSO, and a quoted Enterprise tier. The Continuum app is free on every client, and hosted inference is a separate opt-in plan.

No. Both coordinate parallel agents in isolated worktrees, but Continuum is a multi-device, multi-provider control plane with live quota gauges and a local cost ledger. Conductor is a macOS parallel-workspace product that now sells cloud sandboxes and team collaboration on top. The overlap is worktrees; the product shape diverges after that.

Yes. Plenty of people keep a specialty desktop board and use Continuum for mobile approvals, multi-account gauges, and cross-provider Code sessions. Nothing forces an uninstall, and Conductor's free tier means the pairing costs nothing until you want its cloud hours.

Conductor is a macOS app. Continuum's Mac client is the reference host, but the same session projects to a web client and to beta Windows and Linux desktop builds, and a Continuum agent can enroll a Linux or Windows box as an execution host.

Conductor Cloud runs each agent in an isolated Linux sandbox with your repos and dependencies pre-installed, so work continues after you disconnect. It starts on the $50/mo Pro tier. Continuum's answer is a machine you already control: enroll a Linux or Windows host over Tailscale, or use a Continuum Cloud runner if your account is entitled to one.

Yes. Sessions can run in isolated worktrees with branch-aware UI, live diffs, staged hunks, and PR flows. The product is the whole Code workbench across devices, not only the worktree manager.

Yes. Each session gets its own worktree, and Continuum isolates provider config per account with CLAUDE_CONFIG_DIR and CODEX_HOME, so a second Claude Max or ChatGPT login never clobbers the first. The gauges then show each account's 5h and weekly windows separately.

The spawn grid is an ephemeral 4, 6, or 8 tile grid of agent CLIs for interactive terminal work, and it dies with the app. Conductor's board is a durable orchestration surface for worktree tasks. Different mental models, and some people run both.

Pairing is built around reachable hosts, and Tailscale MagicDNS is the first-class path so an iPhone can reach the Mac daemon without port-forward theater. Loopback still works on the Mac itself, and a Continuum Cloud account can carry the pairing instead.

Further reading

Guide

How to run multiple Claude Code sessions in parallel

Two to six agents at once: the isolation, the rate-limit math, and how to review the output without drowning.

Read guide → Guide

Git worktrees: the complete guide

One repository, many working directories. The mechanics, the full command surface, and the six things that go wrong.

Read guide → Guide

Claude Code rate limits: the 5-hour window and weekly caps explained

Why you got cut off, when it comes back, and the settings that buy the most headroom.

Read guide →
Begin

Run your agents
in Continuum.

Free app. Your subscriptions. Optional hosted inference. Mac stable - Windows and Linux desktop are beta.

Download for Mac See pricing

vendor-neutral · local-first · multi-device