clitipshow-to

10 Clawbrowser CLI Tips You're Probably Not Using

Clawbrowser TeamJune 18, 2025

The Clawbrowser CLI has a short learning curve for basic use. The depth reveals itself over time. Here are ten features worth knowing now.

1. Chain commands with pipes

Extract from multiple pages and pipe results together:

claw extract "product name and price" \
  --url https://shop.example.com/category/keyboards \
  --array --json | \
  jq '[.[] | select(.price < 100)]'

Extract → filter with jq → feed to next step. Standard Unix pipelines work.

2. Use --dry-run to preview actions

Before running a destructive or expensive command, preview what Clawbrowser will do:

claw batch extract "product details" \
  --urls products.txt \
  --dry-run

Output:

DRY RUN — no requests will be made
URLs to process: 247
Estimated time at 30/min: ~8 minutes
Rate limit: 30/minute
Profile: default

No requests made. Just the plan.

3. --verbose for debugging

When extraction isn't working as expected, --verbose shows the full execution trace:

claw extract "price" \
  --url https://shop.example.com/product/123 \
  --verbose

[1] Launching browser context (profile: default)
[2] Navigating to https://shop.example.com/product/123
[3] Waiting for DOM stability...
[4] DOM stable after 1240ms (47 mutations settled)
[5] Running extraction: "price"
[6] Candidate elements found: 3
[7] Best match: .product-price (confidence: 0.94)
[8] Extracted value: "$89.99"

See exactly what Clawbrowser did and why.

4. Named output formats

Control output format for different pipeline needs:

# JSON (default structured)
claw extract "product details" --url ... --json

# CSV (for spreadsheets)
claw extract "name, price, rating" --url ... --csv

# TSV (for some data pipelines)
claw extract "name, price, rating" --url ... --tsv

# Plain text (for LLM input)
claw extract "main article content" --url ... --text

5. Screenshot on failure

When an extraction fails, a screenshot tells you why:

claw extract "price" \
  --url https://shop.example.com/product/123 \
  --screenshot-on-failure ./debug/

If extraction fails, ./debug/failure-1234.png shows you exactly what the browser saw — CAPTCHA, login wall, error page.

6. Profile cloning

Create a new profile pre-populated from an existing one:

# Alice and Bob need similar fingerprints but separate sessions
claw profile clone --from base-profile --to alice
claw profile clone --from base-profile --to bob

# Now they share fingerprint config but have isolated cookies/storage

7. Batch processing with concurrency control

claw batch extract "product name and price" \
  --urls products.txt \
  --json \
  --concurrency 5 \     # 5 parallel browser contexts
  --rate-limit 30/min \ # total across all contexts
  --output results.json

Concurrency and rate limiting are independent. 5 parallel contexts hitting 30 requests/minute total — Clawbrowser distributes the capacity.

8. Wait for specific content

Instead of timeouts, wait for meaningful signals:

# Wait for specific text to appear
claw navigate https://app.example.com/dashboard \
  --wait-for-text "Welcome back"

# Wait for a specific element
claw navigate https://app.example.com/dashboard \
  --wait-for "account balance widget"

# Wait for network to go quiet
claw navigate https://app.example.com/dashboard \
  --wait-for-idle 500ms

Each targets a different readiness signal. Use what fits the page.

9. Environment variables for secrets

Don't put credentials in scripts:

# Set once in your environment
export CLAW_PROXY_URL="http://user:pass@proxy.provider.com:8080"
export CLAW_DEFAULT_PROFILE="production"

# CLI picks them up automatically
claw navigate https://target.com
# Uses $CLAW_DEFAULT_PROFILE and $CLAW_PROXY_URL

Full list of env vars: claw env --list

10. Config files for shared pipeline settings

For repeatable pipelines, put shared settings in .clawrc:

# .clawrc
default_profile: production
rate_limit: 30/minute
delay_jitter: 25%
screenshot_on_failure: ./debug/
output_format: json
timeout: 30s

All commands in the directory pick up these defaults. No flags needed for common settings.

These ten features don't change what Clawbrowser can do — they change how much effort it takes to do it at scale. Worth five minutes to set up, saves hours over a production pipeline's lifetime.

Full CLI reference: claw --help or the docs.

Ready to start?

Give your agent a browser

Clawbrowser handles fingerprints, proxies, and sessions. Your agent focuses on the task.

Get started free

Back to all posts