Why Codealong
Running coding agents in parallel gets messy fast: a dozen terminal tabs, the Claude and Codex
apps, two or three accounts, and no way to tell which agent is waiting on you without clicking
through all of them. Then the pull requests they open pile up, waiting on a review.
Codealong works beside the agents you already run. There's nothing to move into. It finds
every agent session on your machine, wherever it started, shows what each one is doing, and brings
the ones that need you to wherever you are. Keep starting agents in your own terminal and in the
Claude, Codex and Cursor apps: Codealong reads the records each one already writes. It doesn't
wrap, proxy or replace anything, and it needs no account, no API key and no server.
Every agent, account and machine on one board
Every session from every agent and account, wherever you started it: your terminal tabs,
background jobs, the Claude and Codex apps. Sorted into Needs you, Your turn, Working, In
review and Idle, and organized the way your work is:
project → repo → worktree → branch → pull request.
- Read and reply. Any session's whole conversation, with its tool calls, images and links,
searchable, and a box to reply. Or double-click to open it where it lives: its exact terminal
tab, its Claude app session or its Codex thread.
- Know where your plan limits stand. Each account's limits across the top, always in view, and
what your sessions used against what your plans cost.
- All your machines. Link them and each shows the others' sessions beside its own; allow it,
and you answer an agent on your desktop from your laptop. An
iPhone app is on the way.
Keep the agents you already run, where you run them
Apps that host your agents, in their own window or their own terminal panes, see only what you
start inside them. Codealong sees what you already run, and leaves it running where it is.
- Start agents the way you do now. In your own terminal, as a background job, in the Claude,
Codex, Cursor or Antigravity app. Each shows on the board as it starts: nothing has to be
launched from Codealong, and the quick panel is there for when it's quicker.
- The desktop apps are on the board too. A Claude app session or a Codex thread sits beside
your terminal sessions, with its conversation and its status. It takes your reply from
Codealong, and a double-click opens it in its own app.
- Each agent stays its maker's. Its own tools, prompts and permission rules, and its new
features the day they ship. Nothing sits between an agent and its model.
- Nothing to move out of, either. Quit Codealong and every agent in your terminal or its app
carries on, asking you the usual way.
What works with each agent says what Codealong can do with every
one. Where it can't act, it opens the session where it lives.
An agent reviews every pull request, at no extra cost
Turn on automatic reviews and each pull request that asks for your review is read by an agent
first, on the account, model and effort you pick, told what to look for by instructions you can
edit. It leaves a draft review on your own GitHub account, its comments on the lines they're
about, for you to finish and submit; it can approve for you when it finds nothing that needs
fixing. The PR waits on you only once the agent is done, new commits are reviewed again, and your
sign-off is one keystroke from the notch.
- No new bill. A review is an ordinary agent session on the Claude or ChatGPT plan you already
pay for, on whichever account has room. No review bot to pay for by the seat, no API key, no
charge by the token. The Analytics window shows what your sessions would have cost at API prices,
against what your plans cost.
- Nothing to install on GitHub. No GitHub App, no admin to ask, no review service in between:
the agent runs on your machine in your checkout, reads the PR through your own
gh sign-in, and
your code goes only to the model provider you already use.
- Another model reads what one wrote. The reviewer prefers another agent than the one that
wrote the PR: Codex reviews what Claude Code wrote, and the other way round.
- Your own agents' PRs too. Each PR a session of yours opens can be reviewed the same way, and
the review handed back to the session that wrote it to fix, before a teammate looks. Nothing
goes on GitHub.
- Shared with your team. A
.codealong/config.toml committed to the repo carries what every
review looks for, and which verdicts an agent may never submit.
Review with… (⌃⌘R) starts the same review by hand on any PR, and Ask puts questions to an agent
that has read the whole diff. More in
Pull requests, with their agents.
Your notch is an inbox for your agents
Codealong docks in your MacBook's notch. Closed, it counts everything that waits on you: agents
that want to run something, questions, pull requests to review, finished turns. Press ⌥⌘. in
any app and it opens on the agent that has waited longest, with the call it wants to run and
Allow, Always and Deny. ⌘⏎ answers and brings up the next one. You never leave
your editor.
- Approvals, questions and replies. A question comes as choices to pick, with a box for an
answer of your own. Deny a call and say what to do instead in the same keystroke.
- Pull requests too. A PR waiting on your review is in the same queue, with its diff and its
checks: ⌘⏎ approves it.
- No notch needed. An external display gets one drawn for it. Or float the bar over every app,
as it does on Linux, keep it in the menu bar, pin a session to your desktop, or work through
everything in the Waiting on you window.
- Later is a key away. ⌃⌘1 to ⌃⌘7 snooze what can wait, and a short sound says when something
new needs you.
This is the beginning of the README. Read the rest at the original. Read the full README ↗
