Walk through a change out loud — the diff, the terminal, the reasoning — and hand over a recording your teammates can watch and your coding agent can read.
Get the file, the terminal, and the ticket on screen first. Navigation noise at the start of a walkthrough is the part everyone scrubs past, and it is also the part that produces useless key moments.
Say what the change does and why you made that call, then show the diff, the test, and the result. The reasoning is the part that does not survive in a pull request description, which is the whole reason to record instead of type.
Builds, installs, and failing commands are dead air. Pause and resume into the same recording rather than cutting a second clip, so the sequence stays intact for whoever reads it later.
Send the watch link to a person. Give a coding agent the same recording's agent-readable file and it can work from the steps without watching anything.
If the recording is about something broken rather than something built, the bug report recorder is tuned for reproduction steps. More on how agents read recordings is on the agents page.
Every Clipy recording has a second face: a text document containing the transcript, a summary, and a list of key moments with timestamps and what was on screen at each one. An agent can read that in a second or two. It cannot watch a twelve-minute video, and pasting a raw transcript loses everything that was visual. That is the practical difference for a code walkthrough. "It breaks on the second save" is meaningless in a transcript on its own. Tied to a key moment that records what was on screen when the sentence was said, it becomes something an agent can act on. Because this page declares the recording as a walkthrough before you start, the key-moment planner treats it as procedural: it keeps up to 100 moments rather than 30, and is told to preserve every demonstrated action, input, and verification instead of compressing them into a summary. For a walkthrough, a missing step is worse than an extra one.
A PR description is good at what changed and bad at why, because writing out the reasoning takes longer than making the change. The result is a reviewer reconstructing your thinking from the diff, or a round trip of questions. A five-minute walkthrough carries the things a diff cannot: which approach you tried first and abandoned, which part you are least sure about, the failure mode you are guarding against. Link it from the PR and reviewers can take it or leave it. It works the same way in the other direction. A walkthrough recorded while you debug is the fastest way to hand a half-finished investigation to someone in another timezone without writing a report at midnight.
A walkthrough is a sequence, and a summary of a sequence is usually useless — the value is in the order and the completeness. Declaring the recording as a walkthrough before you start means the key-moment planner treats it as procedural from the first pass: up to 100 moments instead of 30, with instructions to preserve every demonstrated action, input, and verification rather than compress them. That is the difference between an agent that can follow your change and one that has the gist of it. The general screen recorder produces the same artifacts and can be re-labelled afterwards. This page just skips the guessing.
A screen recording captures whatever is on the shared surface, including a stray tab, a notification, or an environment file left open in a split pane. Nothing is redacted for you. Share a single window or a single tab rather than your whole display, and close anything holding credentials before you start. The recording is private to your account until you send the link. For anything sensitive, that link can be password-protected or given an expiry date from the recording's settings instead of being the only thing standing in the way.
Yes. Every recording has an agent-readable version containing the transcript, summary, and key moments. Clipy also ships a CLI and an MCP server so an agent can pull it directly rather than being handed a link.
Short enough that someone will watch it — usually under ten minutes. If a change needs longer than that, it is often two recordings, or a sign the change is doing two things.
Yes. Choose the window or the whole display in the browser's sharing prompt. If you record outside a browser often, the Mac app captures a window without a tab in the middle.
No watermark on any recording, on any plan. The free tier has an allowance for how much you can record before upgrading, and the recorder shows you where you are against it.
Tell us what broke. We resolve every report within 24 hours.