Capture Agent
Screenshots an AI can read. Capture on Mac or iPhone, get a link, and annotate what matters — as structured data a model can act on, not coloured arrows it has to guess at.
Capture anywhere
Menu-bar capture on the Mac — area, window, full screen, scrolling, or a recording. On iPhone, share any screenshot from Photos or pick from the Screenshots album. The link is on your clipboard before the upload finishes.
Structured annotations
Draw a box, write what it means. Each callout is stored as a region plus a label and comment — exposed in the share's metadata and rendered as numbered boxes, so an agent reads your intent instead of parsing pixels.
Built for agents
Every share has a text rendition (OCR, transcript, markdown), JSON metadata and the original bytes. Connect the MCP server to Claude and it can list, read, see and annotate your captures directly.
How a link works
/<id>— AI-readable rendition and the annotated image/<id>/metadata— JSON, includingannotations[]/<id>/annotated— the image with numbered callouts drawn on/<id>/raw— the original file