Compare·Provider
Continuum
Multi-agent workbench
VS
Claude Code logo
Claude Code
Official Anthropic coding CLI

Claude Code in Continuum

Same Claude Code quality, because it is the same binary. Continuum adds the peers, the accounts and the instruments.

Claude Code is Anthropic's official agentic coding CLI: plan mode, tools, subagents, and the Pro or Max rails you already pay for. As of August 2026 Anthropic ships it in the terminal, in VS Code and JetBrains, on the web, in a desktop app, in the iOS and Android Claude apps, in Slack and in GitHub Actions. Continuum is not a fork and not a substitute model. It launches the real claude binary over a PTY under your own login, then wraps it in the parts Anthropic does not sell: peer providers in the same sidebar, two Max accounts on one machine, a git worktree and six review panes per session, live 5-hour and weekly gauges, and a local ledger that prices every session by repo and by day.

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

You want Claude Code beside Codex, Cursor and Grok, two or more Claude accounts on one Mac, worktree panes, live quota gauges, a repo-level cost ledger, and a phone client that steers the session running on hardware you own.

Choose Claude Code when

You want Anthropic's own surfaces and nothing else: the terminal, or the web, desktop and mobile clients Anthropic ships, one login, one vendor, no extra workbench to learn.

Snapshot provider · Continuum hosts it
Dimension Continuum Claude Code
Engine identity Hosts the official claude binary over PTY Is the official Claude agent CLI
Auth path Your Anthropic login · no Continuum markup Claude plan login or setup-token
Accounts per machine Several · CLAUDE_CONFIG_DIR isolation One default login per machine
Other providers Codex · Cursor · Grok · Gemini · OpenCode Claude only, by design
Surfaces Mac · web · Win/Linux · iPhone · Watch Terminal · IDE · web · desktop · iOS/Android
Quota gauges Menu-bar 5h + weekly, per account Limits enforced, no fleet gauge
Analytics Local JSONL → $ by repo/provider/day Account usage, not a repo ledger
Mac stable · web · iPhone · Watch · Win/Linux desktop beta · free app
01

Engine in the terminal vs workbench across devices

In Continuum

A Claude day with the workbench open

08:50

The menu-bar gauge shows the 5-hour window still cold. Continuum Code spawns Claude in a fresh worktree for the morning ticket, so main stays clean while the agent works.

13:10

Plan mode finishes while you are out. Approve it from the iPhone. The diff and PR panes stay attached to the same thread when you get back, so nothing has to be re-explained.

16:30

A Codex peer session takes the same ticket in a second worktree. Compare the two diffs, keep the stronger branch, and the loser is archived rather than merged by momentum.

19:20

Primary Max is hot. Pin the secondary account for the night batch. Both accounts roll into one ledger, so the week's dollars by repo still add up in the morning.

In Claude Code

Anthropic's own surfaces only

08:50

Open a terminal and run claude in the repo. Plan mode, tools and subagents work exactly as Anthropic ships them, which is the whole point of the engine.

13:10

Kick a cloud run from the web or the Claude mobile app and let Anthropic host it. Steering the session on your own laptop still means being at that laptop.

16:30

Comparing to Codex means opening a different product. There is no second provider in this pane, and that is a deliberate design choice, not an oversight.

19:20

One machine login. When the weekly window runs hot you wait, or you rearrange accounts by hand. Pure claude stays perfect for quick one-offs with zero extra chrome.

02

Monday path

Monday on Claude Max with Continuum in front

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

01

Install Continuum on the Mac, confirm claude is on PATH, and check that your Pro or Max login already works in a plain terminal.

02

Open the Code tab, start a session on the Claude agent in worktree mode, and let the morning ticket run on its own branch checkout.

03

Watch the menu-bar 5-hour gauge before a heavy run, and the Usage tab for weekly heat. Background: Claude Code limits.

04

Pair the iPhone by QR or Tailscale so a plan that finishes at lunch can be approved without walking back to the desk.

05

Mid-day, open a Codex peer session on the same repo when you want a second line of attack. Claude stays primary; the diffs get compared at review.

06

If the primary window runs hot, pin a secondary Claude account. Continuum isolates it with CLAUDE_CONFIG_DIR instead of swapping HOME, so git, ssh and gh keep working.

07

After the first session, confirm the gauge actually ticks. If it does not, fix auth before stacking a second account on top of a broken one.

08

Evening: read the Usage tab for dollars by repo, then let the night batch run on the secondary account while the primary window recovers.

How Continuum actually runs Claude Code

How Continuum actually runs Claude Code

  • Continuum launches the official claude binary over a real PTY with a subscription-safe environment. No HOME swap, no reimplemented agent, no proxy in front of Anthropic.
  • Plan mode, tool calls and the live transcript stream into the Code panes: chat, plan, diff, PR, terminal and artifacts, all bound to the same session.
  • Secondary accounts get CLAUDE_CONFIG_DIR isolation plus a per-instance Keychain partition, and an unresolvable account pin refuses to spawn rather than silently billing the primary.
  • Gauges poll Anthropic's own rate-limit rails for the 5-hour and weekly windows, per account, in the menu bar and the Usage tab.
  • Analytics walk ~/.claude/projects JSONL, including secondary instance directories, dedupe by message and request id the way ccusage does, and price the result with a LiteLLM snapshot.
  • iPhone and Watch pair to the Mac daemon over Tailscale or MagicDNS for approve, interrupt, diff and PR, with an idempotent outbox so a retried tap never double-sends.
  • A session can run on an enrolled Linux or Windows host instead of the Mac, with the same panes and the same phone control on top.
03

Claude Code host matrix

Same engine under your own login, so this is not a re-score of model quality. Continuum cells name the cockpit: panes, gauges, account isolation and the ledger. Anthropic cells name what Anthropic already ships, including the web, desktop and mobile surfaces it added through 2026.

Capability Continuum Claude Code
Official Claude Code agent Drives the real CLI over PTY Is the CLI
Plan mode and approvePlan never bypasses approvals on either path Plan pane + approve from phone Native plan mode
Subagents and skills The CLI's own, plus a slash palette Native subagents and skills
Codex, Cursor, Grok in the same app First-class peers Claude only
Two Claude accounts on one machinePersonal Max and work Max without a re-login dance CLAUDE_CONFIG_DIR + Keychain split One login per machine
Phone control of a sessionBoth exist; they reach different machines iPhone + Watch steer your host Claude apps drive Anthropic runs
Web and desktop clients Web + Windows/Linux desktop beta Claude Code on web and desktop
Git worktree per session Worktree mode + spawn grid You arrange git yourself
Diff, PR and artifacts panes Hunk staging, gh PR, QuickLook Terminal output and your tools
Live 5h and weekly gauges Menu bar + Usage tab, per account Limits apply, no live gauge
Local $ by repo and provider JSONL priced with LiteLLM rates Not a repo-level ledger
Run on Linux or Windows hosts you enrol Go agent host + same panes CLI runs anywhere you install it
Spawn grid of interactive CLI tiles 4, 6 or 8 agent tiles at once One terminal, one focus
Cost of the workbench itself Free · optional hosted inference Included with your Claude plan
04

What Continuum layers on without forking

01 · Layering

Engine versus cockpit.

Claude Code is the agent. Continuum is the workbench that runs Claude Code and its peers. Treating them as mutually exclusive misses the point: Continuum is how you operationalize Claude Code across devices, accounts and other labs, and its panes attach to the very same session the terminal would have run alone.

02 · Multi-account

Two Max plans, one instrument.

The raw CLI is one login per machine. Continuum isolates secondary accounts with CLAUDE_CONFIG_DIR and per-instance secrets, so you can pin a session to work or personal and flip when a window runs hot. It fails closed on an unresolvable pin rather than quietly billing the primary, which is the failure mode that actually costs money.

03 · Fleet

Claude is not the only good agent.

Some tickets prefer Codex, Cursor or Grok. Continuum keeps Claude Code as a peer rather than a monopoly: open a second session on another provider, compare the diffs, keep the better branch. Claude Code alone stays Claude-only by design, and that design is correct for Anthropic and limiting for you. See Codex vs Claude Code.

04 · Away from the desk

Both have phones. They reach different machines.

Anthropic ships Claude Code on the web and in the iOS and Android apps, so mobile is no longer a Continuum exclusive. What Continuum's iPhone and Watch clients add is control of the session running on hardware you own, with the quota gauges and the spend ledger on the same screen as the approve button.

05 · Cost truth

The plan tells you when you stopped. It does not tell you why.

Anthropic enforces the limits; it does not hand you a repo-level ledger. Continuum parses local session history, dedupes it the way ccusage does, prices it with a LiteLLM snapshot and shows which repo and which provider ate the week. Details in Claude Code cost.

05

Coexistence

Stack recipe

Same engine. Bigger cockpit.

This is not a fork fight. Continuum spawns and steers the real Claude Code binary under your Claude login, so usage bills your Pro or Max plan and Continuum never proxies Anthropic tokens for those sessions. Stack recipe: keep claude naked in a terminal for quick one-offs, use Anthropic's own web and mobile surfaces for Anthropic-managed runs, and open Continuum when the day is multi-account, multi-provider, or happening away from the desk on a host you own.

Download Continuum Providers
06

Host Claude Code or stay CLI-only

Continuum

Choose Continuum to operationalize Claude Code

  • Host the real Claude Code with Mac, web, Windows, Linux, iPhone and Watch clients on top.
  • Stack multiple Claude accounts without clobbering the Keychain login the CLI already uses.
  • Sit Claude beside Codex, Cursor and Grok in one sidebar and compare their diffs.
  • Watch live 5-hour and weekly gauges per account before committing to a heavy run.
  • Read repo-level spend from local history instead of guessing. See usage dashboards.
  • Keep the workbench free on top of the Anthropic plan you already pay for.
Claude Code

Stay on Anthropic's own surfaces when they are enough

  • Want only Anthropic's official surfaces with zero extra UI to learn.
  • Prefer one terminal pane and one Claude login on the machine.
  • Are happy with Anthropic-hosted web and mobile runs for away-from-desk work.
  • Do not need multi-provider comparison or a second account on the same Mac.
  • Already orchestrate git worktrees and parallel sessions with your own scripts.
  • Value the smallest possible tooling surface over fleet ops features.
07

What you pay

Continuum

App + your labs

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

Claude Code

Their bill

Included with your Claude plan as of August 2026: Free (limited), Pro $20/mo or $200 billed yearly, Max 5x $100/mo and Max 20x $200/mo. Team seats are $25/mo billed monthly or $20/mo annual, with a premium seat at $125/mo monthly or $100/mo annual, and Enterprise is custom. Consumption billing on the API is per token. Claude Code draws the same budget as Claude chat: Claude Code pricing.

How to compare

Total cost of work

Continuum does not mark up Claude subscription sessions; it drives the official CLI under your login. The only bill Continuum can add is optional hosted inference, and it is a separate line from Anthropic's. Limits explained in Claude Code limits.

Claude Code 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

Claude Code + Continuum.

Deep dives: docs, providers, sessions.

No, it is a Claude Code host. Continuum runs Anthropic's own binary and adds the peers, accounts, panes and gauges around it. If you only want the CLI with no extra UI, use Claude Code alone. If you want multi-provider, multi-account, mobile and analytics, add Continuum on top.

Yes. As of August 2026 Claude Code is included on Free (limited), Pro at $20/mo or $200/yr, Max 5x at $100/mo, Max 20x at $200/mo, and on Team and Enterprise seats. It draws the same budget as Claude chat, so a heavy coding day shortens your chat window too. See Claude Code pricing.

No. Continuum drives the official CLI under your login, so tokens for those sessions never route through Continuum as a billed proxy. Continuum-hosted inference is a separate, optional product with its own price list.

Anthropic ships Claude Code in its iOS and Android apps, plus web, desktop, Slack and GitHub Actions, so mobile is not a Continuum exclusive. The difference is reach: Continuum's iPhone and Watch clients control the session running on your own Mac or enrolled host, with the plan approve button, the diff and the quota gauge on one screen.

Yes. Pair the iPhone app and you get the same thread, chat, plan, diff and PR, against the session running on your Mac or enrolled host. Sends go through an idempotent outbox, so a retry on a flaky connection cannot double-post a prompt.

Continuum isolates secondary instances with CLAUDE_CONFIG_DIR and per-instance Keychain partitions, never a full HOME override, because swapping HOME breaks git, ssh, gh and node inside worktrees. The primary account keeps the exact environment the bare CLI would have used.

No, and it deliberately does not try. Continuum reimplements no part of Anthropic's agent. Claude Code stays the engine; Continuum is the multi-device, multi-account workbench around it, and it dies gracefully if you uninstall it.

Continuum's Usage tab parses local session history, dedupes it by message and request id, and prices it with a LiteLLM snapshot, giving dollars by repo, provider and day beside the live windows. The method mirrors ccusage: track Claude Code cost.

Code-mode sessions skip approvals by design so an agent run does not stall on a confirm dialog; Plan mode never bypasses. Continuum still surfaces plan approve on the phone. Read the security tradeoff first: AI coding agent security.

Further reading

Guide

Claude Code pricing: Pro vs Max vs API when the weekly cap fires

Claude Code pricing when the weekly cap fires: Pro vs Max vs API, two stacked subscriptions, or Get Plus.

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 → Guide

How to install Claude Code on macOS, Windows, and Linux

The native installer replaced npm as the recommended path. Here is every route, who owns updates on each, and how to fix the common breakages.

Read guide →
Begin

Run Claude Code
from Continuum.

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

Get Plus Download for Mac

vendor-neutral · local-first · multi-device