Observation

`brow snapshot`

Get the accessibility tree of the current page. This is the primary way to read page content.

brow snapshot -s <id> [--search <regex>] [--locator <selector>]

Flag	Description
`--search`	Filter tree to nodes matching this regex
`--locator`	Restrict tree to a subtree rooted at this selector

brow snapshot -s 1                           # full tree
brow snapshot -s 1 --search "price"          # only nodes containing "price"
brow snapshot -s 1 --locator "main"          # only the <main> element subtree

Interactive elements are annotated with refs in brackets — use them with --ref N in click/fill/select:

[1] button "Add to cart"
[2] input "Email address"
[3] a "Read more" href="/article/42"

If the page has more than ~500 interactive elements, the snapshot is truncated and a hint is shown. Use --search to filter.

`brow screenshot`

Capture a screenshot. Saved to ~/.brow/screenshots/ by default.

brow screenshot -s <id> [--full] [--path <file>] [--width <px>] [--scale <f>] [--quality low|medium|high]

Flag	Default	Description
`--full`	`false`	Capture full page (not just viewport)
`--path`	auto-generated	Save path
`--width`	viewport	Resize browser to this width before capture
`--scale`	`1.0`	Scale factor
`--quality`	none	Preset: `low` (400px), `medium` (800px), `high` (1200px)

brow screenshot -s 1                          # viewport screenshot
brow screenshot -s 1 --full                   # full page
brow screenshot -s 1 --quality medium         # 800px wide
brow screenshot -s 1 --path /tmp/page.png     # specific path

Returns the file path of the saved screenshot.

`brow html`

Get the HTML content of the page (or a subtree).

brow html -s <id> [--locator <selector>] [--search <regex>]

brow html -s 1                        # full page HTML
brow html -s 1 --locator "article"   # just the <article> element
brow html -s 1 --search "price"      # lines containing "price"

Use html when you need raw markup that snapshot doesn’t capture (e.g. table data, custom attributes).

`brow logs`

Get console log messages from the page.

brow logs -s <id> [--search <str>] [--count <n>]

Flag	Default	Description
`--search`	none	Filter by substring
`--count`	`50`	Number of recent entries to return

brow logs -s 1
brow logs -s 1 --search "error" --count 100

Captures all console.log, console.warn, console.error, etc. messages.

`brow network`

View HTTP requests made by the browser.

brow network -s <id> [--search <str>] [--count <n>] [--response] [--include-static] [--clear]

Flag	Default	Description
`--search`	none	Filter by URL or content-type
`--count`	`50`	Number of recent entries
`--response`	`false`	Include first 200 chars of response body
`--include-static`	`false`	Include images, fonts, CSS, JS requests
`--clear`	—	Reset the log

brow network -s 1                              # last 50 API requests
brow network -s 1 --search "api"              # filter by URL
brow network -s 1 --response                  # show response previews
brow network -s 1 --clear                     # reset before next navigation

Output format:

POST   200  application/json    https://api.example.com/v2/search
  → {"results":[{"id":1,"name":"foo"...
GET    200  application/json    https://api.example.com/v2/user/profile

This command is the foundation of API scouting. See the API Scouting tutorial.

`brow fetch`

Make an HTTP request using the browser’s session cookies (or without cookies to test auth).

brow fetch -s <id> <url> [-X <method>] [-H <header>] [-d <body>] [--no-cookies]

Flag	Default	Description
`-X` / `--method`	`GET`	HTTP method
`-H` / `--header`	—	Header in `Key: Value` format (repeatable)
`-d` / `--data`	—	Request body
`--no-cookies`	`false`	Skip browser cookies (plain httpx request)

# Authenticated request (with browser cookies)
brow fetch -s 1 "https://api.example.com/user/me" | jq

# Test if endpoint works without auth
brow fetch -s 1 "https://api.example.com/public/data" --no-cookies

# POST with JSON body
brow fetch -s 1 "https://api.example.com/search" \
  -X POST \
  -H "Content-Type: application/json" \
  -d '{"query":"test"}'

The response body is written to stdout. Combine with jq for JSON exploration.

`brow websocket`

View WebSocket messages captured during the session.

brow websocket -s <id> [--search <str>] [--count <n>] [--clear]

Flag	Default	Description
`--search`	none	Filter by message content
`--count`	`50`	Number of recent messages
`--clear`	—	Reset the log

brow websocket -s 1
brow websocket -s 1 --search "patch"      # find JSON Patch messages
brow websocket -s 1 --clear               # reset before next navigation

Output format:

recv  wss://realtime.example.com/socket  {"op":"replace","path":"/odds/1","value":1.85}
sent  wss://realtime.example.com/socket  42["subscribe",{"channel":"matches"}]

WebSocket messages are captured automatically from session start, including socket.io frames.