Let your AI make the video.

Muse, Instinct, ChatGPT and Claude can already read your booking emails. Give them this prompt and they’ll send back a Miles link with your trip ready to play.

1. Copy this prompt into your AI

Look through my email for my most recent trip: flights, trains, buses, ferries, and taxi or Uber rides over 30 km. Put the stops in travel order, then send me one link in exactly this format: /create?r=STOPS&m=MODES&d=DATES&n=TITLE STOPS: comma-separated, airport codes for flights (DEL) and city names otherwise (Agra) MODES: comma-separated, one per leg: flight, train, car, bus or ferry (taxi and Uber count as car) DATES: comma-separated YYYY-MM-DD, one per leg, in departure order TITLE: a short trip name, URL-encoded Example: /create?r=BOM,DEL,Agra,Jaipur&m=flight,train,car&d=2026-11-02,2026-11-04,2026-11-06&n=Golden%20triangle Only include travel I actually took. Leave out anything you're unsure about.

Change “my most recent trip” to any trip you like, such as “my Japan trip in April”.

We never see your email

Your AI reads your inbox; Miles only receives the places, dates and ways you travelled, inside the link. Nothing is stored on our servers.

For developers and agents

Build a link directly. Stops are airport codes for flights or city names otherwise; modes are flight, train, car, bus or ferry; one date per leg.

/create?r=BOM,DEL,Agra,Jaipur&m=flight,train,car&d=2026-11-02,2026-11-04,2026-11-06&n=Golden%20triangle

Or call the API. It returns an edit link, a watch-only share link, a preview image, and any stops it couldn’t match.

curl -X POST https://<this-site>/api/trips \
  -H 'Content-Type: application/json' \
  -d '{"title":"Golden triangle","legs":[
    {"from":"BOM","to":"DEL","mode":"flight","date":"2026-11-02"},
    {"from":"DEL","to":"Agra","mode":"train","date":"2026-11-04"},
    {"from":"Agra","to":"Jaipur","mode":"car"}]}'

Machine-readable guide: /llms.txt · OpenAPI spec: /openapi.json.

Get the video file

POST /api/videos with the same trip renders an MP4 on our server (a few minutes). Poll the returned statusUrl until it’s done, then send the user videoUrl: a 720p video that plays in iMessage and WhatsApp.

MCP server

Agents that speak the Model Context Protocol can connect to /mcp on this site (Streamable HTTP, no sign-in). Tools: render_video returns the MP4, create_trip builds the watch and edit links, find_place checks how a stop will be matched, and list_looks lists the styles. A Muse connector is on the way.