Clipy is a screen recording and context transfer tool for coding agents, positioned as "THE AGENT-READABLE SCREEN RECORDER." Its core idea is: instead of repeatedly describing a problem in words, record a screen video and let coding agents such as Claude Code, Codex, and Cursor directly read the context in the recording, so they understand what changes need to be made.
Use Clipy is used to record bugs, operation demonstrations, or feedback. After the user records once, they paste the link into a connected coding agent, and the agent can read the recording's summary, transcript text, click information, and key-moment frames. The official description emphasizes "type less, show more," giving the agent the full picture.
Main capabilities
- Screen recording: you can choose to record a browser tab, a window, or the entire screen.
- System and microphone audio: you can explain changes in your own voice; system audio can be included when the browser and capture source support it.
- Agent-readable deliverable: Clipy processes the recording into a summary, transcript text, click information, key-moment frames, and pointer context.
- Share link: one link is both for people to watch and for agents to read; it is not publicly listed by default, anyone with the link can open it, and viewers do not need to sign up.
- Flare highlight pointer: in the Mac app or Chrome extension, hold Control and move the pointer to produce a highlighted trail, helping explain "which one you mean." Clipy turns the highlight into a key moment: the frame at the end of the gesture, the pointer position, and what was said at the time. The trail is click-through, and the underlying app still works normally.
- Resolution: supports recording quality from HD to 4K.
- Browser operation: the web version requires no installation, no prompt pop-ups, and no configuration.
How to use
- Choose what to record: a browser tab, window, or screen, showing what happened and what it should become.
- Record and narrate: explain the desired result in your own voice.
- Clipy processes the recording into an agent-readable summary, transcript, and key moments.
- Paste the link into a connected coding agent, keeping both the screen and the explanation while it works. The official text also mentions that agents can record their own proof and demonstrations, and supports "Local AI" to keep things local. The official text does not fully elaborate on the specific operational details.
Applicable scenarios
- Reproducing UI bugs: show the steps, the faulty state, and the expected result, for Claude Code provide evidence investigated in the code repository.
- Walk through UI changes: point out the spacing, behavior, or layout you want changed, and place the explanation together with the relevant screen.
- Explain customer feedback: place the customer's words together with visible examples, let the agent produce a change list, and resolve questions before editing.
- Migrate from Loom: existing Loom recordings can first be brought into the coding workflow, with official free download and usage guidance provided (specific steps are not yet provided).
Limitations and notes
- Free for Mac, Chrome, and web, and viewers do not need to sign up; the official description says "start for free," and signing in with Google allows saving and sharing links. No more detailed pricing or plan information is provided.
- Officially published coding benchmarks show that on a certain synthetic UI task, structured recording context outperforms input with transcript text only; the report includes the recording, context, scoring criteria, and the results of Codex's 20 main trials. The official statement clearly notes that this does not mean the same improvement will occur on every task, model, or competing tool.
- Links are unlisted by default, and anyone with the link can open them.
- For some capabilities (such as system audio, Local AI, and the agent recording proof itself), the official text does not fully provide the specific scope of support and steps.
