# Continuum: Claude, Codex, Grok — Unified AI chat, code & usage

> Unified AI chat, code & usage (iPhone/iPad app by Montauk Analytics Inc..)

- Source: https://appshunter.io/ios/app/continuum-claude-codex-grok/id6776332528 (this page in markdown: same URL + `.md`)
- Developer: [Montauk Analytics Inc.](https://appshunter.io/developer/6776332530)
- Category: Developer Tools, Utilities
- Price: Free
- Age rating: 4+
- Requires: iOS 26.0 · 108 MB
- Languages: American English
- Released: 2026-06-09
- Data updated: 2026-08-22
- User reviews in markdown: https://appshunter.io/ios/app/continuum-claude-codex-grok/id6776332528/reviews.md

## What is Continuum?

Continuum is your unified workspace for AI chat, coding agents, and usage analytics — on iPhone, Apple Watch, and Mac, kept in sync.

CHAT — one place for every model
Talk to Claude, ChatGPT, Gemini, Cursor, Grok, and OpenRouter from a single chat. Broadcast one prompt to several models at once, compare their answers side by side, then continue with the one you like best.

CODE — drive your coding agents from anywhere
Start, monitor, and steer Claude Code, Codex, and other CLI coding agents running on your Mac — right from your iPhone or Apple Watch. Review diffs and pull requests, approve plans, watch live terminals, and reply by voice. Your phone becomes a remote control for your dev machine.

ANALYTICS — never get surprised by a limit
Live rate-limit gauges for Claude Code and Codex show how much of your 5-hour and weekly limits you've used. Historical token and cost analytics break spend down by project and model, with Apple Watch complications that put it all on your wrist.

WHY CONTINUUM
• Three tools in one — chat, code, and analytics, instead of juggling apps.
• Pairs with your Mac — the free Continuum Mac app does the heavy lifting and securely streams to your iPhone and Watch over your own network. No account required.
• Private by design — your data stays on your devices. Nothing is collected.
• Built for people who live in Claude Code, Codex, and the terminal.

Pair once with the Continuum Mac app and your chats, sessions, and usage follow you to every screen.


## Version history (last 5 releases)

### 1.0.18 — 2026-08-01

# Continuum 0.72.0

Follow-up to 0.71.0's sync work: your Mac stays reachable on its own, and the
phone finally shows the truth about what's running.

## What's new

- **Your Mac heals itself instead of going dark.** After a crash-relaunch or a
  stale enrollment, the Mac could get stuck "offline" forever — no restart or
  re-sign-in brought it back. It now detects the stall, clears the bad
  enrollment, and re-enrolls itself automatically. (#1352)
- **PR controls on your phone finally know what state you're in.** Instead of a
  permanent orange **Create PR** button — which stayed loud after the work was
  done and offered to open a second PR against an already-merged branch — iOS
  and Android now show the same three-state control the Mac has: *Create PR*
  when there isn't one, a colour-coded PR pill when there is, and a *Merge*
  button when checks are green. (#1367)
- **Usage gauges work again on iPhone.** They were pinned at 0% with a
  recurring "Your Mac didn't respond in time" — the phone was asking the wrong
  machine. It now reads usage from a live Mac, and holds the last good numbers
  through a relay blip instead of flickering every five seconds. (#1365, #1366)
- **Live sessions actually look live.** A session running on another device
  showed as finished — no data-stream animation, no "Live on another device".
  Running sessions now report their real status across every client. (#1365,
  #1366)

## Under the hood

- Host-control startup gets a 30-second watchdog and retries on the mirror
  timer, so a single failed enroll can no longer strand a machine. (#1352)
- Sessions resumed after an app restart re-attach their event wiring, so the
  session registry stops publishing a frozen row. (#1366)
- PR-state vocabulary moved into one shared model (Swift + TypeScript) with
  matching tests, so the Mac, iOS, and Expo clients can't drift apart on what a
  merged PR looks like. (#1367)

### 1.0.13 — 2026-07-22

# Continuum 0.65.5

A Figma-style comment-and-edit loop right inside the Code Preview, plus smoother Code pane
resizing on Mac.

## New

- **Comment and edit on the live Preview.** The Code review Browser tab is now **Preview**:
  localhost, GitHub, and `.local` links open in-app instead of Safari, and you can drop
  Claude Design–style comment bubbles on any element or switch to a Figma-like text-edit
  mode. Selected elements and text edits become first-class chips in the composer and
  transcript, so the agent gets the exact selector and context you pointed at. (#1251)

## Improved

- **Smoother Code pane resizing.** Dragging the sidebar and review-pane dividers now tracks
  the cursor cleanly — no rubber-banding or animation fight — and the width sticks after you
  let go. (#1251)

### 1.0.12 — 2026-07-21

# Continuum 0.65.2

Continuum Plus arrives — a $25/month tier with hosted inference built in — alongside four
free hosted models in the picker, a rebuilt first-run experience, and a round of Chat, Code,
and billing fixes across iOS and Mac.

## New

- **Continuum Plus.** A new $25/month plan that includes hosted inference with a $25/week
  usage cap, so you can run models through Continuum without bringing your own key. The native
  app shows a live weekly usage gauge, and the cloud dashboard surfaces the tier label and
  remaining budget. (#1230, #1231, #1232)
- **Four free hosted models.** The Continuum picker now offers Nemotron 3 Ultra, Nemotron 3
  Super, Gemma 4 31B, and North Mini Code at no cost — pick one and start coding without a
  paid provider. The hosted catalog was also trimmed to only models that actually serve, so
  you never select one that can't run. (#1238, #1235)
- **A rebuilt first run.** Onboarding is now a guided, per-provider setup with cleaner empty
  states and a working Quick Start — no more confusing web-org step. New installs get to a
  working session faster. (#1228)

## Fixed

- **Model selection that tells you the truth.** Choosing Continuum as your provider now
  surfaces gateway errors instead of failing silently, only offers models the live catalog can
  actually serve, and counts Continuum correctly in broadcast. (#1227)
- **No more hidden voice-shortcut capture on Mac.** The voice shortcut no longer grabs
  keystrokes when it shouldn't. (#1226)
- **Cleaner model menu on Mac.** Settings no longer repeats the "Continuum" brand prefix on
  every hosted model name. (#1225)
- **Fable sends reliably.** Recovered a first-send idempotency conflict that could stall the
  very first message to a Fable model. (#1234)
- **Accurate document tabs on Mac.** Tool output with a bare filename no longer fabricates a
  bogus document-tab path. (#1240)
- **iOS Chat & Code polish.** The New Chat button keeps its intrinsic size, Code session-card
  diff stats are vertically centered, and the header CTAs are more compact with the Code
  "Add a device" promo removed. (#1222, #1223, #1224)
- **Correct Fable pricing.** claude-fable-5 is now billed at its list rate ($10/$50 per Mtok)
  instead of falling back to a Sonnet-class price. (#1233)
- **Admin invites, fixed.** Creating an invite no longer throws a silent 520, "Pro" is renamed
  to "Continuum App," the email field is sized correctly, and invites now default to a 3-month
  Continuum App code. (#1229, #1239)

### 1.0.11 — 2026-07-20

# Continuum 0.65.1

Mesh connections that stay connected, a monochrome Continuum mark that fits the tray, and a
batch of dogfood polish across Chat, Code, and Settings on iOS and Mac.

## Fixed

- **Your devices stay connected.** iOS mesh connections now persist and recover on their own
  instead of getting stuck "handshaking" — no more re-selecting a device and hoping it
  connects. A watchdog gives up on a dead rendezvous and reconnects cleanly. (#1219)
- **Every device is selectable.** The Devices list is now grouped into Available and Other,
  with your Macs first and hosts sorted by session count. Windows and Linux hosts show as
  connected and can run sessions directly — pick any available device from New Session. (#1219)
- **A Continuum mark that fits.** The Continuum provider glyph is now a monochrome mark that
  matches the app icon, sitting alongside Codex, Cursor, and Grok in the model tray instead of
  a colored chip. (#1219)
- **Chat, Code, and Settings polish.** Full Cursor model catalog in New Session, Continuum
  hosted inference shows only when you've enabled it, Code rows carry Mac-parity git branch/PR
  icons, and a round of Chat/Settings/Usage cleanup from dogfooding. (#1221)

## Under the hood

- **Security headers restored on the web.** continuumcode.ai now sends clickjacking and HSTS
  protection (X-Frame-Options, CSP frame-ancestors, Strict-Transport-Security) on every
  response — these were dropped when the backend moved to the self-hosted box. (#1220)
- **Web release points at the live backend.** `release-web.sh` now forwards to the box deploy
  instead of the retired Cloudflare Worker path. (#1220)

### 1.0.8 — 2026-07-14

# Continuum 0.62.0

Spend numbers you can trust: token costs now price Anthropic's five-minute and one-hour cache writes separately — with the rate that was actually in effect when the usage happened — and the Mac, Continuum Cloud, and the web dashboard all agree to the cent. Chat gets a big one too: Broadcast can now **Synthesize** the replies from every provider into one cited answer. Plus clickable links, configurable App Shot trigger keys, and a batch of transcript fixes.

## Features

- **Broadcast → Synthesize.** After every selected provider finishes replying, one click compiles their answers into a single cited response — each claim traceable to the exact reply it came from — using the summarizer model you picked in the composer. Continue from the synthesis and it becomes a normal solo chat with just your question and the answer. (#1086)
- **Pick your App Shot trigger keys.** The both-modifiers screenshot chord is now configurable: both Options (default), both Commands, or both Shifts — switchable in Settings and applied live, with no accidental fires while another modifier is held. (#1091)
- **Links in chat are actually clickable.** PR links and bare URLs in replies open in your browser on click, show the hand cursor, and reveal the full destination on hover. Only http/https/mailto links are ever openable. (#1088)

## Fixed

- **Token spend is now correct across every surface.** Anthropic 5-minute vs 1-hour cache writes are priced at their real rates (1.25× vs 2×), historical usage uses the rate active at the time (including Sonnet 5's launch promotion, which expires Sept 1), OpenAI cached input is validated before discounting, and a corrupt pricing download can no longer wipe known rates. Mac analytics, Continuum Cloud, and the web dashboard now compute the identical dollar figure. (#1092)
- **Background sessions keep their activity.** Switching a session between providers (Claude → Codex/Grok/Cursor) no longer risks losing chat activity or leaking a stale feed into the new provider's transcript. (#1093)
- **Slash-skill messages show what you typed — once.** Invoking a skill now renders one user bubble with your exact `/skill arguments` text and one SKILL.md attachment, instead of duplicate provider payloads and stray markdown attachments. Two invocations of the same skill no longer merge into one. (#1090)
- **Finished sub-agents say Done.** Synchronous Agent runs that returned their full report no longer sit on "Started" with a running timer forever; only genuinely backgrounded agents stay on Started until they finish. (#1089)
- **Interrupted turns show how long they ran.** A turn you stop (or that errors) now displays its elapsed time in the footer, even though there's no final answer to copy. (#1085)
- **Slash-skill search ranks by name match.** Typing `/rev` puts `review` above skills that merely mention "review" in their description — exact name, then prefix, then contains, then description-only. (#1087)
- **Less noise beside the live timer.** The redundant "thinking…" caption next to the active-turn data stream is gone; the stream and timer are the working-state cue. (#1084)

## Under the hood

- Xcode project generation is now byte-deterministic across worktrees (a checkout-named group was churning the project file on every machine), and duplicate idempotency-key handling releases its reservation before an early-exit response so a fast retry can't see a phantom "in flight" conflict. (#1091, #1086)

## Apps similar to Continuum

| App | Rating | Price | Category |
| --- | --- | --- | --- |
| [VibeBuddy: Agent Monitor](https://appshunter.io/ios/app/vibebuddy-agent-monitor/id6777469338) | — | Free | Developer Tools |
| [DNP Remote](https://appshunter.io/ios/app/dnp-remote/id6760618642) | — | Free | Developer Tools |
| [Codex Meter](https://appshunter.io/ios/app/codex-meter/id6781059780) | — | Free | Developer Tools |
| [Ruflet Explorer](https://appshunter.io/ios/app/ruflet-explorer/id6762528151) | — | Free | Developer Tools |
| [Pathrule](https://appshunter.io/ios/app/pathrule/id6786861213) | — | Free | Developer Tools |
| [Agent Tick](https://appshunter.io/ios/app/agent-tick/id6765610411) | — | Free | Developer Tools |

## More apps by Montauk Analytics Inc.

- [Glide - Personal & Business](https://appshunter.io/ios/app/glide-personal-and-business/id6776708031)
- [PG Essays](https://appshunter.io/ios/app/pg-essays/id6780459369)

All apps by Montauk Analytics Inc.: https://appshunter.io/developer/6776332530

## Related topics

[manifold continuum care](https://appshunter.io/ios/topics/manifold-continuum-care) · [grok 4](https://appshunter.io/ios/topics/grok-4)

---

*Data collected daily from the US App Store and indexed by [AppsHunter](https://appshunter.io/). User reviews are verbatim App Store reviews. Ratings, prices and chart positions refresh continuously; this snapshot is from 2026-08-22.*