Replays cost nothing
The player starts no agent, types nothing into a terminal and sends nothing to a model. Watch a long run again tomorrow, or a month later, for free.
Workspace recording
Record the whole squad’s run.Replay it byte for byte.
Start a recording and NeuroSquad keeps everything the workspace shows: every byte each terminal printed, every status, the notes, the browser frames. Later you replay it on a timeline — at any speed, as often as you like, without running an agent or sending a single request to a model.
- Every terminal byte, every agent
- Replays cost no tokens
- Stays on your computer
A recording of the recipes workspace replays in a copy of NeuroSquad’s player. A note called Task is sent to two connected agents at once. Claude Code reads the recipes page, writes a filter, adds a search box and checks it in the browser card by typing a query; the dev server’s terminal logs each reload. Codex writes tests, asks for approval to install vitest, gets it, and runs the tests until nine pass. Both add a line to the note and finish. The timeline has markers for the prompt being sent, Codex needing input and each agent finishing; you can play, pause, change the speed from 1× to 64×, jump between markers, scrub, and switch to the clean view.
Claude CodeWorking
TaskRecipe searchClaude Code: search by name and ingredient, query in the URLCodex: tests for the filter, then run them
CodexWorking
dev server
BrowserRecipes3 recipesbasilBasil pesto pastaTomato basil soupThai basil stir-fry
Markers
Everything here is replayed from one recording: scrub anywhere, change the speed, jump between markers — the terminals land on exactly the same screen every time.
Why record
A run you can watch again, as often as you need
Agents take minutes or hours, and what they did scrolls away. A recording keeps the whole workspace on one timeline.
Terminals, byte for byte
Every byte each terminal printed is replayed into a real terminal, colours, cursor moves and resizes included. An automated test compares a live terminal with its replay at checkpoints — seeking forward, backward and at random — and they match.
Find the moment fast
Play at 1× to 64×, drag the timeline, or jump between markers: when a prompt was sent, when an agent needed you, when each one finished. ← and → step five seconds.
Shoot the video later
Clean view hides every control and shows the canvas at exactly the size your live canvas had, panned and zoomed the way you left it. Point any screen recorder at it — at the speed you choose.
See what each agent really did
Every key and paste sent to a terminal, every status change, the notes and to-do lists as they changed, what the browser cards showed. Compare two agents on the same task after the fact, or attach the run to a bug report.
Yours, and it survives a crash
Recordings are written as they happen and flushed to disk every few seconds, on your computer only. If the app or the machine goes down, the recording is marked Recovered and plays up to where it stopped.
What goes in
Everything the workspace shows, in one timeline
One recording covers one workspace, from Start to Stop, as a single ordered log.
- Terminals
Every byte of output, starts, exits and resizes — of agent CLIs and plain shells alike, so all 19 CLIs are covered.
- Input
Every key, paste and prompt sent to a terminal, including the prompts the app sends for you.
- Statuses
Working, needs input, finished — each change, with the moment it happened.
- The canvas
Cards, groups and arrows, and the window’s pan and zoom, sampled once a second.
- Notes and to-dos
Their text, each time it changed — yours and the agents’.
- Browser cards
The frames the card showed, up to four a second, while it was on screen.
- Other cards
Run Stats, Agent Pulse, community cards: a picture about once a second while visible.
- Markers
Prompts sent, needs input, finished — and marks a script adds through the app.
How it works
Start, stop, replay
- 01
RECCanvas tools → Start recording
A red Recording chip with the elapsed time appears in the title bar and stays while it runs.
- 02
Work as usual
Recording changes nothing about how agents and cards behave.
- 03
Stop
Press the red chip or Canvas tools → Stop recording. Quitting NeuroSquad stops it too.
- 04
Canvas tools → Recordings… → Replay
The list shows each recording’s length, size and cards. Replay opens the player over the whole window.
Straight answers
What it keeps, and what it doesn’t
A recording contains everything the workspace showed and every key typed into its terminals — including a password typed into a shell. Treat it like the screen it came from.
Not recorded
- Hover, selection, scrolling inside a card, menus, tooltips, toasts and the mouse pointer.
- Browser frames and card pictures only while the card was on screen and the canvas wasn’t zoomed out to the overview — exactly what you saw.
- A terminal already running when you press Start begins from its current screen and a few hundred lines of scrollback, not its whole history.
- Grid mode is replayed as the canvas; in the player every card wears the shared card header.
Privacy
- Recordings stay on your computer. NeuroSquad never uploads them.
- Keys you enter in NeuroSquad’s own dialogs — provider keys, SSH passwords — are not terminal input and are not recorded.
- The red chip is visible the whole time; recording can’t be started from your phone.
- Delete a recording from Recordings…, or delete its folder.
Workspace recording, answered
Does replaying run the agents again?
No. Nothing in the player is live: no agent is started, nothing is typed into a terminal and nothing is sent to a model. You watch what was recorded.
Can I export a video?
Not as a file from the app. Switch the player to Clean view — the canvas at your live canvas’s exact size, no controls — and record it with any screen recorder, at whatever speed you choose.
Which agents can I record?
All of them. The recording keeps what each terminal printed, not one CLI’s log, so every agent CLI and every plain shell on the canvas replays the same way.
Where are recordings kept, and how big do they get?
In a recordings folder inside NeuroSquad’s data folder, or the folder the NEUROSQUAD_RECORDINGS_DIR environment variable names. Terminal output is small; browser frames take most of the space. Once a recording passes 5 GB, or the disk has less than 2 GB free, browser frames and card pictures pause while terminals keep recording.
What happens if NeuroSquad crashes mid-recording?
You lose at most the last few seconds. The next time you open Recordings…, the cut-off recording is marked Recovered and plays up to where it stopped.
Can I record several workspaces at once?
One recording runs at a time, of one workspace.
Record your next run
Canvas tools → Start recording. Replay it whenever you want, as often as you want.
Windows 10 / 11 · macOS 12 or newer, Apple silicon and Intel