gem-hunter
Language: TypeScript
Official Bright Data CLI - scrape, search, and extract structured web data directly from your terminal.
Scrape, search, and extract structured web data — directly from your terminal.
@brightdata/cli is the official npm package for the Bright Data CLI. It installs the brightdata command (with bdata as a shorthand alias) for access to the full Bright Data API surface:
| Command | What it does |
|---|---|
brightdata scrape |
Scrape any URL — bypasses CAPTCHAs, JS rendering, anti-bot protections |
brightdata search |
Google / Bing / Yandex search with structured JSON output |
brightdata discover |
AI-powered web discovery - find and rank results by intent with optional full-page content |
brightdata scraper create |
Build a Bright Data scraper from a natural-language description using AI |
brightdata scraper run |
Run a Bright Data scraper on a URL and return the data |
brightdata scraper heal |
Fix an existing scraper in place via AI self-healing (stops at an approval gate) |
brightdata scraper approve |
Approve (or reject) a self-healing fix that is awaiting approval |
brightdata pipelines |
Search and run any of thousands of pre-built scrapers (Amazon, LinkedIn, TikTok…) for structured data |
brightdata browser |
Control a real browser via Bright Data's Scraping Browser — navigate, snapshot, click, type, and more |
brightdata zones |
List and inspect your Bright Data proxy zones |
brightdata budget |
View account balance and per-zone cost & bandwidth |
brightdata skill |
Install Bright Data AI agent skills into your coding agent |
brightdata add mcp |
Add the Bright Data MCP server to Claude Code, Cursor, or Codex |
brightdata config |
Manage CLI configuration |
brightdata init |
Interactive setup wizard |
Requires Node.js ≥ 20
curl -fsSL https://cli.brightdata.com/install.sh | shnpm install -g @brightdata/cli
npm install -g @brightdata/cli
You can also run without installing:
npx --yes --package @brightdata/cli brightdata <command>
# 1. Run the interactive setup wizard brightdata init # 2. Scrape a page as markdown brightdata scrape https://example.com # 3. Search Google brightdata search "web scraping best practices" # 4. Find a scraper and get structured data brightdata pipelines search linkedin profiles # 5. Check your account balance brightdata budget # 6. Install the Bright Data MCP server into your coding agent brightdata add mcp
Get your API key from brightdata.com/cp/setting/users.
# Interactive — opens browser, saves key automatically brightdata login # Via GitHub CLI — no browser needed (requires gh) brightdata login --github # Non-interactive — pass key directly brightdata login --api-key <your-api-key> # Environment variable — no login required export BRIGHTDATA_API_KEY=your-api-key
On first login the CLI checks for required zones (cli_unlocker, cli_browser) and creates them automatically if missing.
# Clear saved credentials brightdata logout
brightdata add mcp uses the API key stored by brightdata login. It does not currently read BRIGHTDATA_API_KEY or the global --api-key flag, so log in first before using it.
Every new Bright Data account includes a recurring monthly free tier — no credit card or commitment required to start. See the Free Tier docs for full details.
5,000 free credits per month (≈ $7.50), renewing on the 1st of each month. Unused credits don't roll over.
These credits cover the products most CLI commands use:
| Product | CLI commands | Credit cost |
|---|---|---|
| Unlocker API | scrape |
1 credit / request |
| SERP API | search |
1 credit / request |
| Web Scraper API | pipelines |
1 credit / request |
| Scraper Studio | scraper create / run / heal |
1 credit / page load (shared pool, not per record) |
Not included in the monthly free credits: proxy products (Datacenter, ISP, Residential) and the Browser API (brightdata browser). These get a separate one-time $2 trial (valid 7 days), plus a $5 bonus when you add a payment method (valid 30 days).
Adding a credit card is a verification step only — you are not charged unless your free credits are exhausted and you have funds deposited. Accounts on custom pay-as-you-go or pre-commit subscription plans don't receive the recurring monthly free credits.
Check your remaining balance any time with brightdata budget.
Interactive setup wizard. The recommended way to get started.
brightdata init
Walks through: API key detection → zone selection → default output format → quick-start examples.
| Flag | Description |
|---|---|
--skip-auth |
Skip the authentication step |
-k, --api-key <key> |
Provide API key directly |
Scrape any URL using Bright Data's Web Unlocker. Handles CAPTCHAs, JavaScript rendering, and anti-bot protections automatically.
brightdata scrape <url> [options]
| Flag | Description |
|---|---|
-f, --format <fmt> |
markdown · html · screenshot · json (default: markdown) |
--country <code> |
Geo-target by ISO country code (e.g. us, de, jp) |
--zone <name> |
Web Unlocker zone name |
--mobile |
Use a mobile user agent |
--async |
Submit async job, return a response ID |
-o, --output <path> |
Write output to file |
--json / --pretty |
JSON output (raw / indented) |
-k, --api-key <key> |
Override API key |
Examples
# Scrape as markdown (default) brightdata scrape https://news.ycombinator.com # Scrape as raw HTML brightdata scrape https://example.com -f html # US geo-targeting, save to file brightdata scrape https://amazon.com -f json --country us -o product.json # Pipe to a markdown viewer brightdata scrape https://docs.github.com | glow - # Async — returns a response ID brightdata scrape https://example.com --async
Search Google, Bing, or Yandex via Bright Data's SERP API. Google results include structured data (organic results, ads, people-also-ask, related searches).
brightdata search <query> [options]
| Flag | Description |
|---|---|
--engine <name> |
google · bing · yandex (default: google) |
--country <code> |
Localized results (e.g. us, de) |
--language <code> |
Language code (e.g. en, fr) |
--page <n> |
Page number, 0-indexed (default: 0) |
--type <type> |
web · news · images · shopping (default: web) |
--device <type> |
desktop · mobile |
--zone <name> |
SERP zone name |
-o, --output <path> |
Write output to file |
--json / --pretty |
JSON output (raw / indented) |
-k, --api-key <key> |
Override API key |
Examples
# Formatted table output (default) brightdata search "typescript best practices" # German localized results brightdata search "restaurants berlin" --country de --language de # News search brightdata search "AI regulation" --type news # Page 2 of results brightdata search "web scraping" --page 1 # Extract just the URLs brightdata search "open source scraping" --json | jq -r '.organic[].link' # Search Bing brightdata search "bright data pricing" --engine bing
AI-powered web discovery. Submit a query with optional intent, and Bright Data finds, ranks, and optionally extracts full-page content for each result.
brightdata discover <query> [options]
| Flag | Description |
|---|---|
--intent <text> |
AI intent to evaluate and rank result relevance |
--country <code> |
ISO country code (default: US) |
--city <name> |
City for localized results (e.g. "New York") |
--language <code> |
Language code (default: en) |
--num-results <n> |
Number of results to return |
--filter-keywords <words> |
Comma-separated keywords that must appear in results |
--include-content |
Include full page content in each result |
--no-remove-duplicates |
Keep duplicate results |
--start-date <date> |
Only content updated from date (YYYY-MM-DD) |
--end-date <date> |
Only content updated until date (YYYY-MM-DD) |
--timeout <seconds> |
Polling timeout (default: 600) |
-o, --output <path> |
Write output to file |
--json / --pretty |
JSON output (raw / indented) |
-k, --api-key <key> |
Override API key |
Examples
# Basic discovery — table output brightdata discover "AI trends" # With AI intent for relevance ranking brightdata discover "AI trends" \ --intent "Prioritize institutional reports for VC research" # Include full page content as markdown brightdata discover "AI trends" --include-content --num-results 5 # Geo-targeted with date range brightdata discover "best restaurants" --country US --city "New York" \ --start-date 2025-01-01 --end-date 2025-12-31 # Filter results by keywords brightdata discover "generative AI SaaS" --filter-keywords "revenue,SaaS" # JSON output to file brightdata discover "AI trends" --num-results 10 --pretty -o results.json # Pipe-friendly — redirected stdout outputs JSON automatically brightdata discover "AI trends" --include-content --num-results 3 > results.json
Build a Bright Data scraper from a natural-language description using AI.
brightdata scraper create <url> <description> [options]
| Flag | Description |
|---|---|
--name <name> |
Scraper template name (default: cli-scraper-<timestamp>) |
--deliver-webhook <url> |
Webhook URL for the deliver stub (default: https://example.com/webhook) |
--timeout <seconds> |
Polling timeout in seconds (default: 600) |
--max-retries <n> |
Max retries on the AI-Flow concurrent-job cap 429 (default: 4). See below. |
--no-retry |
Fail immediately on 429 instead of waiting. Same as --max-retries 0. |
-o, --output <path> |
Write the JSON envelope to a file (see below) |
--json / --pretty |
JSON output (raw / indented) |
--legacy-output |
Write the pre-v0.3 bare AI-progress payload to -o instead of the envelope. Migration only. |
--timing |
Show request timing |
-k, --api-key <key> |
Override API key |
Note: The scraper is created with a placeholder webhook delivery target (
https://example.com/webhook). You can reconfigure the actual delivery endpoint in the Bright Data web UI after creation.
Every termination path — success or failure — writes the same JSON envelope shape:
{
"collector_id": "c_mp7x8a9b2c0d1e2f",
"name": "my-product-scraper",
"status": "done",
"completed_steps": ["prepare_intent_analyzer", "planner", "..."],
"view_url": "https://brightdata.com/cp/scrapers/c_mp7x8a9b2c0d1e2f",
"created_at": "2026-05-18T07:28:30Z"
}On failure paths the envelope adds an error field and the status reflects the failure category (ai_trigger_failed, failed, poll_failed). The collector_id and view_url are still present so you can recover or inspect the half-built scraper.
This makes the documented chain in recipes.md work as written:
brightdata scraper create https://example.com/product/1 "..." \ -o create.json COLLECTOR_ID=$(jq -r '.collector_id' create.json) brightdata scraper run "$COLLECTOR_ID" https://example.com/product/2
The file format follows the
-oextension, so.jsonis written compact (ideal forjq). Use--prettyfor indented JSON on stdout when you omit-o.
Use --legacy-output if you have an existing script that depended on the pre-v0.3 bare-progress shape; the flag is supported for one minor version while you migrate.
The Bright Data AI Flow caps concurrent scraper create generations per account (currently 3). If you exceed it, the API returns 429 Cannot run more than N jobs in parallel. The CLI handles this automatically: it waits with exponential backoff + jitter and retries up to --max-retries times (default 4). During the wait the CLI prints status lines so you know it isn't hung:
Triggering AI generation...
Hit AI-Flow concurrent-job cap (429). Waiting 32s before retry 1/4...
Hit AI-Flow concurrent-job cap (429). Waiting 67s before retry 2/4...
Generating scraper...
If the cap is still hit after all retries, the CLI exits with a stderr note pointing at the half-built collector's dashboard URL so you can inspect or delete it manually (Bright Data does not yet expose programmatic collector deletion).
Use --no-retry if you want the old fail-fast behavior — typically for scripts that prefer to handle backoff themselves.
Examples
# Build a scraper for a product page brightdata scraper create https://example.com/product/1 \ "Extract title, price, and image URL from this product page" # Name the scraper and save the envelope to a file brightdata scraper create https://example.com/product/1 \ "Extract title, price, and image URL from this product page" \ --name my-product-scraper -o create.json # Capture the collector_id for chaining COLLECTOR_ID=$(jq -r '.collector_id' create.json) # Fan out 10 parallel creates — the CLI serialises automatically via 429 backoff for url in $(cat urls.txt); do brightdata scraper create "$url" "Extract title, price, ..." \ --name "scraper-$(basename $url)" & done; wait # Disable the auto-backoff (fail fast on 429) brightdata scraper create https://example.com/product/1 \ "Extract title, price, and image URL from this product page" \ --no-retry # Use a custom webhook delivery URL brightdata scraper create https://example.com/product/1 \ "Extract title, price, and image URL from this product page" \ --deliver-webhook https://my-app.com/ingest
Run a scraper (built with scraper create or in the web UI) against one or more URLs and get the extracted data.
brightdata scraper run <collector_id> [url] [options]
Provide URLs in exactly one of three ways:
<url> — single URL (legacy form, unchanged).--urls <u1,u2,...> — comma-separated list.--input-file <path> — file with one URL per line, or a JSON array of URL strings, or a JSON array of {"url": "..."} objects.| Flag | Description |
|---|---|
--urls <list> |
Comma-separated list of URLs (multi-URL batch path) |
--input-file <path> |
File with URLs (txt one-per-line, or JSON array) |
--sync |
Use the synchronous /dca/crawl endpoint (single-URL only, server-side cap of 25–50s) |
--sync-timeout <seconds> |
Sync-mode server timeout, 25–50 (default: 50) |
--timeout <seconds> |
Polling timeout (default: 600 single-URL, 3600 batch) |
--name <name> |
Human-readable job name |
--version <version> |
Scraper version (e.g. dev) |
-o, --output <path> |
Write output to file |
--json / --pretty |
JSON output (raw / indented) |
--timing |
Show request timing |
-k, --api-key <key> |
Override API key |
Routing
--urls / --input-file) → async flow: /dca/trigger_immediate → poll /dca/get_result. Use --sync for /dca/crawl (one-shot, 25–50s).--urls / --input-file with 2+ entries) → single POST to /dca/trigger with an array body, one collection_id, polled via /dca/dataset. This mirrors the canonical batch shape used by the reference SDKs (triggerWithUrls / trigger_with_urls). --sync is incompatible with multi-URL — /dca/crawl accepts only a single URL.If a single URL expands to more pages than the realtime job limit allows (paginated listings, infinite scroll), the CLI automatically falls back to the batch endpoint and prints a one-line notice. No flag required.
Examples
# Default: async + poll until results arrive brightdata scraper run c_mp3tuab31lswoxvpws https://www.amazon.com/dp/B08N5WRWNW # Save pretty-printed results to a file brightdata scraper run c_mp3tuab31lswoxvpws https://www.amazon.com/dp/B08N5WRWNW \ --pretty -o product.json # Sync mode for fast pages brightdata scraper run c_mp3tuab31lswoxvpws https://example.com/p/1 --sync # Sync with a shorter server timeout and a job name brightdata scraper run c_mp3tuab31lswoxvpws https://example.com/p/1 \ --sync --sync-timeout 30 --name first-test # Multi-URL batch — one API call, one snapshot, one merged result array brightdata scraper run c_mp3tuab31lswoxvpws \ --urls "https://example.com/p/1,https://example.com/p/2,https://example.com/p/3" \ --pretty -o products.json # Multi-URL from a file (one URL per line; # comments and blank lines skipped) brightdata scraper run c_mp3tuab31lswoxvpws --input-file urls.txt -o products.json # Multi-URL from a JSON array echo '["https://example.com/p/1","https://example.com/p/2"]' > urls.json brightdata scraper run c_mp3tuab31lswoxvpws --input-file urls.json
Fix an existing scraper in place when it ran but returned wrong, empty, or partial data. The collector_id stays the same — the scraper is improved, not replaced. This is the maintenance twin of scraper create: it triggers Bright Data's AI self-healing flow (POST /dca/collectors/{id}/refactor_template), then polls progress.
brightdata scraper heal <collector_id> "<prompt>" [options]
You are the detector. The CLI never decides on its own that a scraper is broken — you inspect the run output and decide. The <prompt> is required (max 1000 chars); name exactly what is wrong and what the correct output should be. Vague prompts produce vague heals.
| Flag | Description |
|---|---|
--url <url> |
Verify target woven into the success next_step hint (not sent to the heal call) |
--auto-approve |
When the heal hits the approval gate, approve it automatically and poll through to done (default: stop and let you review) |
--timeout <seconds> |
Polling timeout (default: 600) |
--max-retries <n> |
Max retries on the AI-Flow concurrent-job-cap 429 (default: 4) |
--no-retry |
Fail immediately on 429 instead of waiting through the cap |
-o, --output <path> |
Write output to file |
--json / --pretty |
JSON output (raw / indented) |
--legacy-output |
Emit the bare AI-progress payload instead of the envelope |
--timing |
Show request timing |
-k, --api-key <key> |
Override API key |
The approval gate
Self-healing is human-in-the-loop. Without --auto-approve, heal runs the fix and then stops at an approval gate rather than committing it, exiting 0 with a status: "awaiting_approval" envelope:
{
"collector_id": "c_mp3tuab31lswoxvpws",
"status": "awaiting_approval",
"prompt": "Price returns null — the selector moved …",
"preview_result": [ { "title": "…", "price": { "value": 51.77, "currency": "GBP" } }, … ],
"diff_summary": "proposed template has 1 step(s) — review at view_url",
"view_url": "https://brightdata.com/cp/scrapers/c_mp3tuab31lswoxvpws",
"next_step": "bdata scraper approve c_mp3tuab31lswoxvpws --url https://example.com/product/1"
}preview_result shows the sample rows the fixed scraper would produce — review them, then run the next_step (scraper approve) to commit. awaiting_approval is not a failure; it means the fix is ready and waiting for your decision. A failed heal (429 cap exhausted, timeout, terminal failed) is non-destructive — the existing scraper is unchanged and still works as before.
Examples
# Heal a scraper, stop at the gate, and get a ready-to-run verify command back brightdata scraper heal c_mp3tuab31lswoxvpws \ "The price field returns null — the selector moved into a span with \ data-testid. Capture price and currency again." \ --url https://example.com/product/1 --pretty -o heal.json # Fully autonomous: heal and approve in one command (no manual review) brightdata scraper heal c_mp3tuab31lswoxvpws \ "Reviews stopped extracting after the page redesign" --auto-approve
Commit (or reject) a self-healing fix that scraper heal left awaiting approval. Calls POST /dca/collectors/{id}/resume_automation_job, then polls the refactor job to done.
brightdata scraper approve <collector_id> [options]
| Flag | Description |
|---|---|
--reject |
Reject the proposed fix instead of approving it |
--url <url> |
Verify target woven into the success next_step hint |
--timeout <seconds> |
Polling timeout (default: 600) |
-o, --output <path> |
Write output to file |
--json / --pretty |
JSON output (raw / indented) |
--legacy-output |
Emit the bare AI-progress payload instead of the envelope |
--timing |
Show request timing |
-k, --api-key <key> |
Override API key |
On success the job advances to status: "done" and the envelope hands back a next_step = scraper run <id> <url> so you can verify the committed fix. --reject discards the proposed fix (status: "rejected") — re-run scraper heal with a sharper prompt to try again. If a heal needs multiple approvals, approve may stop at awaiting_approval again — just run it once more.
The self-healing loop
# 1. Run and inspect the data brightdata scraper run c_mp3tuab31lswoxvpws https://example.com/product/1 --json -o out.json # 2. If the data is wrong, heal (stops at the approval gate) brightdata scraper heal c_mp3tuab31lswoxvpws \ "Price returns null — the selector moved; capture price + currency." \ --url https://example.com/product/1 --pretty -o heal.json # 3. Review heal.json's preview_result, then approve brightdata scraper approve c_mp3tuab31lswoxvpws \ --url https://example.com/product/1 --pretty -o approve.json # 4. Verify the committed fix brightdata scraper run c_mp3tuab31lswoxvpws https://example.com/product/1 --pretty
Search and run any of thousands of pre-built Scraper API scrapers and get structured data back. The scraper list comes from Bright Data's docs and is cached for 24 hours.
brightdata pipelines search <query...> # find a scraper brightdata pipelines info <dataset_id> [method] # its methods, or one method's input fields brightdata pipelines run <dataset_id> <method> [urls...] [--<field> <value>...] brightdata pipelines status <snapshot_id> [--wait] # progress of a run brightdata pipelines results <snapshot_id> # records of a finished run brightdata pipelines refresh # reload the scraper list now
run takes the scraper's input fields as flags, one --<field> <value> per field listed by info. Values are converted by field type (number, true/false, comma-separated arrays). Each URL argument becomes one input record. Unknown or missing required fields stop the command before anything is triggered or billed. Use --input for different values per record or nested data.
Flag (run) |
Description |
|---|---|
--<field> <value> |
An input field of the method (see info) |
--input <json|@file> |
Input records as JSON, or @path to a JSON file |
--limit-per-input <n> |
discover_by_* only: max records per input (default: 10). Every record is billed |
--async |
Print the snapshot ID instead of waiting |
--format <fmt> |
json · csv · ndjson · jsonl (default: json) |
--timeout <seconds> |
How long to wait (default: 600) |
-o, --output <path> |
Write output to file |
--json / --pretty |
JSON output (raw / indented) |
-k, --api-key <key> |
Override API key |
Examples
# 1. Find a scraper brightdata pipelines search amazon reviews # 2. See its input fields, an example and a ready-to-run command brightdata pipelines info gd_le8e811kzy4ggddlq collect_by_url # 3. Run it on a URL brightdata pipelines run gd_le8e811kzy4ggddlq collect_by_url \ "https://www.amazon.com/dp/B09V3KXJPB" --max_reviews 20 --pretty # Input fields as flags brightdata pipelines run gd_lwdb4vjm1ehb499uxs collect_by_url \ --keyword laptop --url https://www.amazon.com # Discover method, capped at 5 records per input brightdata pipelines run gd_l7q7dkf244hwjntr0 discover_by_keyword \ --keyword "usb c hub" --limit-per-input 5 # Several records from a file → CSV brightdata pipelines run gd_l7q7dkf244hwjntr0 collect_by_url \ --input @urls.json --format csv -o products.csv # Start without waiting, then check and fetch brightdata pipelines run gd_le8e811kzy4ggddlq collect_by_url "<url>" --async brightdata pipelines status s_abc123xyz --wait brightdata pipelines results s_abc123xyz --pretty
Records with an error field explain inputs that failed, such as a URL of the wrong type.
Changed in 0.4.0: the 43 fixed pipeline types (
pipelines amazon_product <url>,pipelines list) and the top-levelbrightdata statuscommand were removed. Old commands print the replacement. To migrate, find the scraper withbrightdata pipelines search <site or data type>and run it withpipelines run.
Control a real browser session powered by Bright Data's Scraping Browser. A lightweight local daemon holds the browser connection open between commands, giving you persistent state without reconnecting on every call.
brightdata browser open <url> # Start a session and navigate brightdata browser snapshot # Get an accessibility tree of the page brightdata browser screenshot [path] # Take a PNG screenshot brightdata browser click <ref> # Click an element brightdata browser type <ref> <text> # Type into an element brightdata browser fill <ref> <value> # Fill a form field brightdata browser select <ref> <value> # Select a dropdown option brightdata browser check <ref> # Check a checkbox / radio brightdata browser uncheck <ref> # Uncheck a checkbox brightdata browser hover <ref> # Hover over an element brightdata browser scroll # Scroll the page brightdata browser get text [selector] # Get text content brightdata browser get html [selector] # Get HTML content brightdata browser back # Navigate back brightdata browser forward # Navigate forward brightdata browser reload # Reload the page brightdata browser network # Show captured network requests brightdata browser cookies # Show cookies brightdata browser status # Show session state brightdata browser sessions # List all active sessions brightdata browser close # Close session and stop daemon
Global flags (work with every subcommand)
| Flag | Description |
|---|---|
--session <name> |
Session name — run multiple isolated sessions in parallel (default: default) |
--country <code> |
Geo-target by ISO country code (e.g. us, de). On open, changing country reconnects the browser |
--zone <name> |
Scraping Browser zone (default: cli_browser) |
--timeout <ms> |
IPC command timeout in milliseconds (default: 30000) |
--idle-timeout <ms> |
Daemon auto-shutdown after idle (default: 600000 = 10 min) |
--json / --pretty |
JSON output |
-o, --output <path> |
Write output to file |
-k, --api-key <key> |
Override API key |
Navigate to a URL. Starts the daemon and browser session automatically if not already running.
brightdata browser open https://example.com brightdata browser open https://amazon.com --country us --session shop
| Flag | Description |
|---|---|
--country <code> |
Geo-targeting. Reconnects the browser if the country changes on an existing session |
--zone <name> |
Browser zone name |
--idle-timeout <ms> |
Daemon idle timeout for this session |
Capture the page as a text accessibility tree. This is the primary way AI agents read page content — far more token-efficient than raw HTML.
brightdata browser snapshot brightdata browser snapshot --compact # Interactive elements + ancestors only brightdata browser snapshot --interactive # Interactive elements as a flat list brightdata browser snapshot --depth 3 # Limit tree depth brightdata browser snapshot --selector "main" # Scope to a CSS subtree brightdata browser snapshot --wrap # Wrap output in AI-safe content boundaries
Output format:
Page: Example Domain
URL: https://example.com
- heading "Example Domain" [level=1]
- paragraph "This domain is for use in illustrative examples."
- link "More information..." [ref=e1]
Each interactive element gets a ref (e.g. e1, e2) that you pass to click, type, fill, etc.
| Flag | Description |
|---|---|
--compact |
Only interactive elements and their ancestors (70–90% fewer tokens) |
--interactive |
Only interactive elements, as a flat list |
--depth <n> |
Limit tree depth to a non-negative integer |
--selector <sel> |
Scope snapshot to elements matching a CSS selector |
--wrap |
Wrap output in --- BRIGHTDATA_BROWSER_CONTENT ... --- boundaries (useful for AI agent prompt injection safety) |
Capture a PNG screenshot of the current viewport.
brightdata browser screenshot brightdata browser screenshot ./result.png brightdata browser screenshot --full-page -o page.png brightdata browser screenshot --base64
| Flag | Description |
|---|---|
[path] |
Where to save the PNG (default: temp directory) |
--full-page |
Capture the full scrollable page, not just the viewport |
--base64 |
Output base64-encoded PNG data instead of saving to a file |
Click an element by its snapshot ref.
brightdata browser click e3 brightdata browser click e3 --session shop
Type text into an element. Clears the field first by default.
brightdata browser type e5 "search query" brightdata browser type e5 " more text" --append # Append to existing value brightdata browser type e5 "search query" --submit # Press Enter after typing
| Flag | Description |
|---|---|
--append |
Append to existing value using key-by-key simulation |
--submit |
Press Enter after typing |
Fill a form field directly (no keyboard simulation). Use type if you need to trigger keydown/keyup events.
brightdata browser fill e2 "user@example.com"Select a dropdown option by its visible label.
brightdata browser select e4 "United States"
| |
| |
|
| |
| |
| |
|
| |
| |
| |
| |
|
| |
| |
|
| | |
| | |
|
| |
| |
| |
|
| |
| |
|
| |
| |
| |
| |
| |
|
| ||
| ||
|
| |
| |
| |
|