Codealong
One window for every coding agent you run across accounts and machines, with approvals, questions and PR reviews a keystroke away in any app.
- Source repo
- codealong/codealong
- Stars
- ★ 72
- Last updated
- today
- License
- Apache-2.0
- Primary language
- Rust
- FA score
- Insufficient evidence
At a glance
- Works with
- Portable with changesChatGPT · Codex · Claude Code · Claude.ai
- You'll need
- Typical use
- Developers juggling a dozen terminal tabs of Claude Code and Codex who keep missing permission prompts while switching windows.
- Main limitation
- Linux has no installer and must be built from source under X11, including Zig 0.17.0 for the Ghostty terminal library; Windows is only an early source build.
- Website
- codealong.com ↗ set on GitHub
- Maintenance
- Commits in the last 30 dayslast commit 2026-11-20latest release v0.27.0 (2026-11-20)
- Source review
- Insufficient evidence 8 safety controls not found
What does this agent do, and when should you use it?
Codealong is a free, open-source Tauri desktop app for macOS and Linux that collects every coding-agent session on your machine into one board. Rather than wrapping or proxying Claude Code, Codex, OpenCode, Antigravity or Cursor, it reads the records those agents already write and organizes their sessions into projects, repos, worktrees, branches and pull requests. Its main surfaces are a session board with full conversations and reply boxes, a notch or floating bar that queues approvals, questions and PR reviews, an optional agent-run pull-request review, todos and schedules, and end-to-end encrypted links between your machines over Tailscale or iroh. It needs no Codealong account, API key or server, but it depends on the agents and subscriptions you already have.
On startup Codealong scans agent records such as ~/.claude and ~/.codex (plus OpenCode, Antigravity and Cursor) to discover sessions wherever they started, then shows them in the Sessions view sorted into Needs you, Your turn, Working, In review, Merged and Idle. Selecting a session renders its whole conversation with tool calls and images, and the reply box writes back through the agent's own channel: typing into its terminal tab, Claude Code's background service, a new Codex app turn, or OpenCode's service. On macOS the app docks in the notch; ⌥⌘. opens the Waiting on you queue, where Allow/Always/Deny answer command approvals and choices or free text answer questions, with ⌘⏎ advancing to the next waiting agent. The PRs view joins pull requests to the worktrees and sessions behind them and supports review, approve and merge; with automatic reviews on, an agent reads the diff on a chosen account, model and effort and leaves a draft review on your own GitHub account. A codealong CLI and matching skill let agents inspect other sessions, pick up todos, check account usage and start sessions; Schedules starts or messages sessions on cron-like expressions or external events; Settings writes config.toml (~/Library/Application Support/Codealong/config.toml on macOS, ~/.config/codealong/config.toml on Linux), and a repo's .codealong/config.toml shares review policy with a team.
- Developers juggling a dozen terminal tabs of Claude Code and Codex who keep missing permission prompts while switching windows.
- People with several Claude or ChatGPT accounts who want new work routed to whichever account still has plan room.
- Teams that want every PR requesting their review read first by an agent on a different model, leaving a draft review to finish.
- Users moving between a laptop and a desktop who want to read and reply to sessions on the other machine from one board.
- Engineering teams that hand a todo list to agents and wake sessions automatically when CI fails or a PR merges.
Key setup facts
Pulled from the install and usage notes below. The project README has the current values.
- Install / run
brew install --cask codealong/tap/codealong
- Config files
.codealong/config.toml- Runtime
- Rust 1.96+ · Node.js 22.13+
- Install notes cover
- Windows · macOS · Linux
How do you install or deploy this agent?
On macOS, install with Homebrew (this also puts the codealong command on your PATH):
brew install --cask codealong/tap/codealongAlternatively download Codealong.dmg from codealong.com, drag it into Applications and open it; the build is signed and notarized and updates itself.
Linux has no installer yet and must be built from source under X11, with Rust 1.96+, Node 22.13+ (not 23) and Zig 0.17.0 (used to build the Ghostty terminal library). On Ubuntu 24.04, install Tauri's system packages first:
sudo apt update
sudo apt install -y build-essential curl wget file git pkg-config libssl-dev \
libwebkit2gtk-4.1-dev libgtk-3-dev libayatana-appindicator3-dev \
librsvg2-dev libxdo-devThen clone and build:
git clone https://github.com/codealong/codealong.git
cd codealong
npm ci
npm run tauri build -- --bundles debRequirements: macOS 13 or later (Apple silicon or Intel) or Linux under X11 with tmux and a terminal emulator; at least one coding agent installed; and gh auth login for pull-request features. Windows is only an early from-source build.
How do you use this agent?
On first launch Codealong discovers existing sessions and opens the Set up Codealong wizard, where you tick the accounts to show, your code folders, terminal and editor; its last step installs the codealong command, Codealong's plugin for each Claude Code account, and a skill for the other agents. Keep starting agents as you normally do in your terminal or in the Claude, Codex or Cursor apps.
Press ⌥⌘. in any app to work through waiting approvals and questions, or ⌥⌘/ to find sessions, todos and PRs, or to type a prompt and pick the repo, worktree, account, model and place to run it:
codealong whoami # which session am I, who else works here
codealong sessions find "rate limits"
codealong todo list --ready
codealong accounts --usage
codealong start --todo <id>To turn on automatic PR reviews, enable [reviews] auto under Settings › Automation and optionally set the account, model and effort; repository-wide policy lives in .codealong/config.toml and can be checked with codealong config check --repo <folder>. Cross-machine use requires enabling Settings › Devices and pairing once over a Tailscale network or with a link code.
What are this agent's strengths and limitations?
- Nothing has to be migrated: agents keep running in your own terminal or their official apps, and Codealong only reads the records they already write.
- Covers Claude Code, Codex, OpenCode, Antigravity and Cursor, and merges multiple accounts and machines into a single board.
- Automatic PR reviews run on the Claude or ChatGPT plan you already pay for, with no GitHub App, no admin approval and no per-seat review bot.
- The notch and floating bar turn command approvals, question answers and PR approvals into keyboard actions from any app.
- Machine linking is end-to-end encrypted over Tailscale or iroh with no Codealong server in between, and
.codealong/config.tomlshares review policy with a team.
- Linux has no installer and must be built from source under X11, including Zig 0.17.0 for the Ghostty terminal library; Windows is only an early source build.
- Agent support is uneven: Codex and Cursor approvals only work in some terminals, Antigravity and Cursor app sessions open back in their own app, and Linux lacks macOS-only features such as the notch and the Claude/Codex apps.
- It reads local agent records such as ~/.claude and ~/.codex and relies on signed-in agent CLIs plus the GitHub CLI, so it is tightly coupled to your machine setup.
- Several features (boards, canvas, voice, orchestrator, decision models) are marked experimental and must be enabled in Settings › Experimental.
- Automatic reviews and cross-machine control require understanding config.toml account routing, routing groups and review-submission policy.
How does this agent compare with similar options?
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| Codealong This agent | Insufficient evidence | — | ★ 72 | today | Rust | ChatGPT · Codex · Claude Code · Claude.ai |
| Yardsort | 83 · Good | Desktop appFree | ★ 14 | 1d ago | Rust | Codex · Claude Code |
| PandaCode | 67 · Some gaps | — | ★ 1.3k | today | TypeScript | Codex · Claude Code · Claude.ai |
| Harness | 64 · Some gaps | Desktop appFree + model costs | ★ 1.5k | today | TypeScript | Codex · Claude Code |
How does FollowAgents rate this agent?
Why each dimension lost points
Insufficient evidence to score: the packet contains README, AGENTS.md/CLAUDE.md, several SKILL.md files, fragments of main.rs/lib.rs, package.json, workflows and LICENSE, but no reviewable permission declarations, confirmation-flow implementation, data-flow code, or dependency lockfiles. The README claims 'no account, no API key and no server' while also describing cross-device linking, iroh networking, a launchd-resident server, and reading credential-bearing directories such as ~/.claude — a tension that only code could settle, and the code is not present. All trust sub-criteria are scored 0 as absence of evidence, not as proven defects.
Not assessable: no test files, error-handling implementation, or lockfiles (Cargo.lock/package-lock) are in the packet. AGENTS.md asserts tests and a scripts/land.sh retry mechanism, but those are assertions without the corresponding files. Scored 0 throughout.
The SKILL.md descriptions state fairly specific trigger conditions (todos, cross-session messaging, meeting transcripts, account selection), but there is no implementation or boundary documentation to check against; environment fit rests only on README prose about macOS/Linux/Windows. Evidence is insufficient, scored 0.
Documentation structure (README, docs/guide.md, docs/development.md, SECURITY.md, references to CONTRIBUTING.md), the full Apache-2.0 LICENSE text, package.json version 0.26.0 and marketplace.json version are positive signals; however changelog, maintenance ownership and publisher identity are not established in the packet (publisher unverified), and most items rest on README self-description. Too thin to score item by item, so conservatively 0.
The README describes a clear product value (unified board, automatic PR review, notch inbox), but there is no runnable artifact, output sample, or cost data to check; the cost-benefit claim is only the assertion of 'no new bill'. Scored 0.
Every claim comes from README and skill-document self-description; there is no second source inside the packet (no tests, no CI results, no release artifacts). AGENTS.md mentions a history rewrite with preserved refs, a traceability hint that cannot be confirmed statically here. Scored 0.
- Not found in source: least-privilege scopingGrant only what the task needs: a dedicated account or read-only token, scoped to specific directories and repos.
- Not found in source: confirmation before actingTurn on (or add) a confirmation step before it acts, and try it in a sandbox or test environment before real data.
- Not found in source: data-flow disclosureWatch which external services it contacts (proxy or firewall logs) and keep sensitive data out until you know where it goes.
- Not found in source: sensitive-data handlingUse dedicated, low-privilege, revocable API keys — never production credentials — and keep secrets out of logs.
- Not found in source: dependency securityPin versions and run a dependency audit (npm audit, pip-audit) before installing; prefer running it in a container.
- Not found in source: disclosed external effectsEstablish which external systems it writes to, sends to or changes, and verify with test accounts or repos before production.
- Not found in source: rollback or recovery pathBack up first, or work on a git branch or snapshot, so its changes can be undone.
- Not found in source: verifiable attributionInstall from the official repo or registry and check the publisher and URL to avoid look-alike packages.
- The packet is documentation plus a few source fragments; it lacks permission, data-flow, dependency-lock and test evidence, so no substantive safety or reliability judgment is possible.
- The README claims 'no server, no account' while also describing cross-device linking, iroh networking and a launchd-resident server; actual network and data flows must be verified in the full source.
- The app reads agent credential directories such as ~/.claude and ~/.codex and may answer permission prompts on the user's behalf — a high-sensitivity surface whose storage and transport need review.
- Publisher identity is unverified; the update path (latest.json, Homebrew cask) and signing mechanism need independent confirmation.
- AGENTS.md describes agents pushing directly to main and auto-relaunching the app; if misread by an external agent as product instructions this is risky — confirm it is contributor guidance only.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
v0.27.0 Codealong 0.27.0
Release notes excerpt
Every panel in the main window gets the strip a session's has, with Next in each view; decisions can go to OpenAI's or Google's models; your PRs that are ready to merge wait on you; and conversations are easier to scan. What's new - One strip over every panel. A todo's panel has Start, Done, Snooze and Due… along its top, a worktree's New session, Terminal, Editor and GitHub, and a schedule's Run now, Pause and Edit, each with its key. Next (⌘]) and ⌘[ work in every view: Sessions goes through the sessions that wait on you, PRs through its reviews, Todo and Kanban through the todos du …View on GitHub ↗ -
v0.26.0 Codealong 0.26.0
Release notes excerpt
You choose what waits on you and in what order, todos that are due join it in the notch, Claude Code's own dialogs can be answered from Codealong, and the main window gets a calmer look with its status bar at the foot. What's new - Choose what waits on you, and in what order. Settings › Queue (or [queue] in config.toml) lists what the notch, the waiting window and a session's panel go through: sessions that need you, PRs waiting on your review, todos that are due, sessions whose turn ended and todos ready to start. Drag them into your order, or take reviews and todos out, and choose …View on GitHub ↗ -
v0.25.0 Codealong 0.25.0
Release notes excerpt
A session can coordinate work across helpers, sessions carry on once their account is signed in again, PRs an agent is at work on get sections of their own, and each machine says how busy it is. What's new - Coordinate a piece of work across helpers. Hand a session some work and /codealong:coordinate has it plan the parts, start a helper for each (in a worktree of its own when it changes code), keep them going as they report back, and bring you what they come to, asking you only the decisions that are yours. Helpers that finish close together report in one message. …View on GitHub ↗
FAQ
Does Codealong replace my Claude Code or Codex setup?
Do automatic PR reviews cost extra or need a GitHub App?
gh sign-in and leaving a draft review on your own GitHub account; there is no paid review bot and no GitHub App for an admin to install.Does cross-machine linking send my data through a third-party server?
Can I use it on Linux?
What happens for an agent Codealong cannot answer?
Is Codealong open source?
Is Codealong still maintained?
Who makes Codealong?
What is Codealong's official website?
Thanks — we use this to improve the page.