AREC (Agent Recording)

Updated: July 2026

AREC (Agent Recording) is the open markdown format that encodes a screen recording as text an AI agent reads instead of watching video.

AREC defines the fields the document carries and the rules a consumer follows to read it: metadata, summary, key moments with click targets, transcript, and optional cursor paths. It is markdown-first, so a language model reads it with no parser and no SDK. It is the format; an agent-readable recording is the category it encodes.

For example, when you record the Export CSV bug on a payouts dashboard, Clipy produces an AREC document. Paste the share link into Claude Code and the agent reads the summary, the key moment where the button sticks, the click target at the button coordinates, and the console TypeError from exportWorker.ts:142. No video decoding happens on the agent's side.

The document an AREC recording produces is the markdown twin. Public share links resolve to it by appending .md to any watch URL.

Related terms: agent-readable recording, markdown twin, click target.

Learn more: what is an agent-readable recording. Read the format itself in the AREC spec.