Downloads

Launch with "downloads": true and every file the browser downloads in that session is kept. After the session ends the files are listed in the dashboard and under GET /v1/me/downloads, and each one can be fetched through a private link for 2 days.

Two ways to get a file

In your code, while the session runs — no option needed: Playwright's download.saveAs() (or createReadStream()) streams the file to your machine over the session's connection, byte-exact (case 12). Without downloads the file is deleted with the session.
Kept, after the session — with "downloads": true: for downloads your code can't catch, such as a person clicking in live view or Launch interactive, an AI agent working through the MCP server, a client that disconnected (keep_alive) or crashed, or exports you collect later. Your code still gets its downloads as usual.

Launch, then fetch

// 1. Launch with downloads kept
POST https://starter.cdpfleet.com/chrome/session
{ "proxy": "http://user:pass@your-proxy:8080", "headless": "new", "downloads": true }

// 2. After the session ends (files are listed within a few seconds)
GET https://cdpfleet.com/v1/me/downloads?session_id=<sessionId>
→ { "files": [{ "id": "…", "name": "report.pdf", "bytes": 48213, "sha256": "…", "expires_at": "…" }] }

// 3. A private link to one file, valid 15 minutes
GET https://cdpfleet.com/v1/me/downloads/<id>/url
→ { "url": "https://cdpfleet.com/v1/downloads/…", "expires_in_s": 900 }

Every file is served as an attachment of type application/octet-stream: what a web page hands a browser can be anything, so it never opens on cdpfleet.com. Check the sha256 if you pass files on.

Limits

Plans	Every paid plan; not Community (`403 downloads_not_available`)
Per session	Up to 1 GB and 200 files; further files aren't kept (your code still receives them)
Per account	1 GB; when a new session's files would pass it, the oldest files are deleted
Kept for	2 days, then deleted; `DELETE /v1/me/downloads/{id}` deletes one sooner
Browsers	All of them

Not kept

Downloads in browser contexts a raw-CDP client creates itself (Puppeteer's createBrowserContext() over cdpUrl); downloads in the default context and in Playwright contexts are kept.
Contexts opened with acceptDownloads: false: the browser doesn't download at all.
WebKit doesn't treat a text/plain response with Content-Disposition: attachment as a download.