name	description
browser-fetch	Fetch authenticated web content using browser session (SSO, internal sites)

browser-fetch Skill

Fetch web content from authenticated/internal sites using a real browser with persistent login sessions.

Repository: https://github.com/benthomasson/browser-fetch

When to Use This Skill

Use this skill when:

• Fetching internal company pages that require SSO login
• curl returns login pages or 401/403 errors
• WebFetch fails due to authentication
• You need to access content behind Red Hat SSO or similar

Quick Start (Server Mode - Recommended)

# User starts server (token required by default):
uvx --from "git+https://github.com/benthomasson/browser-fetch" browser-fetch --serve https://internal.redhat.com

# Server prints a token. User provides token to Claude. Claude fetches via curl:
curl 'http://localhost:8080/fetch?token=TOKEN&url=https://internal.redhat.com/page&text=true'

IMPORTANT: Ask the user for the token printed by the server.

Alternative: Single-Fetch Mode

# First time: Login to save session (opens browser)
uvx --from "git+https://github.com/benthomasson/browser-fetch" browser-fetch --login https://internal.redhat.com

# After login: Fetch headlessly
uvx --from "git+https://github.com/benthomasson/browser-fetch" browser-fetch --headless --text https://internal.redhat.com/page

Common Commands

Login (First Time)

# Open browser for manual SSO login
uvx --from "git+https://github.com/benthomasson/browser-fetch" browser-fetch --login https://source.redhat.com

# Log in via browser, then press Enter
# Session saved to ~/.config/browser-fetch/profile

Fetch Content

# Fetch full HTML
uvx --from "git+https://github.com/benthomasson/browser-fetch" browser-fetch --headless https://internal.redhat.com/page

# Fetch text only (no HTML tags) - easier to read
uvx --from "git+https://github.com/benthomasson/browser-fetch" browser-fetch --headless --text https://internal.redhat.com/page

# Extract specific element by CSS selector
uvx --from "git+https://github.com/benthomasson/browser-fetch" browser-fetch --headless --selector "main" --text https://docs.example.com/page

# Save to file
uvx --from "git+https://github.com/benthomasson/browser-fetch" browser-fetch --headless --text -o /tmp/page.txt https://internal.redhat.com/page

Options

Option	Description
--login	Open browser for manual login, save session
--headless	Run without visible browser (requires saved session)
--text	Extract text content only (no HTML)
--selector CSS	Extract specific element by CSS selector
-o FILE	Save output to file
--wait N	Wait N seconds after page load (default: 5)
--timeout N	Page load timeout in seconds (default: 30)
--profile-dir DIR	Custom profile directory
--serve	Run as HTTP server (browser stays open)
--port N	Server port (default: 8080)
--no-token-dangerous	Disable security token (NOT recommended)

Server Mode (Recommended for SSO)

User starts server in a separate terminal:

uvx --from "git+https://github.com/benthomasson/browser-fetch" browser-fetch --serve https://source.redhat.com

Server prints a security token. Ask user to provide this token.

Claude fetches via curl (with token):

# Fetch text content
curl 'http://localhost:8080/fetch?token=TOKEN&url=https://source.redhat.com/page&text=true'

# Fetch with selector
curl 'http://localhost:8080/fetch?token=TOKEN&url=https://source.redhat.com/page&selector=main&text=true'

# Check if server is running (no token needed)
curl 'http://localhost:8080/health'

Query parameters:

• token - security token (required by default)
• url - URL to fetch (required)
• text=true - extract text only
• selector=CSS - extract specific element
• wait=N - wait N seconds after load

Workflow Example (Server Mode)

When user asks to read an internal page:

1. Check if server is running:

   curl -s 'http://localhost:8080/health' || echo "Server not running"

2. If not running, ask user to start it and provide token:

   Please start browser-fetch server and give me the token:
   uvx --from "git+https://github.com/benthomasson/browser-fetch" browser-fetch --serve https://source.redhat.com
   

3. If running but you get 401, ask user for the token:

   I got 401 Unauthorized. Please provide the security token from the browser-fetch server output.
   

4. Fetch content via curl (with token):

   curl -s 'http://localhost:8080/fetch?token=USER_PROVIDED_TOKEN&url=https://source.redhat.com/page&text=true'

5. Or save to file and read:

   curl -s 'http://localhost:8080/fetch?token=TOKEN&url=https://source.redhat.com/page&text=true' > /tmp/page.txt

   Then use the Read tool to read /tmp/page.txt

Tips

1. Token is required by default - Prevents other processes from using your session
2. Ask user for token - Token is printed when server starts
3. Server mode for SSO - Keeps browser open, handles auth better
4. Use text=true for readable output without HTML clutter
5. Check health first - curl localhost:8080/health to verify server is up (no token needed)

Comparison with Other Tools

Tool	Use Case
curl	Public pages, APIs with tokens
WebFetch	Public pages (can be slow/unreliable)
browser-fetch	Authenticated pages, SSO, internal sites

Troubleshooting

"Session expired" - Re-run with --login to refresh

"Playwright not installed" - First time setup:

cd ~/git/browser-fetch && uv sync && uv run playwright install chromium

Page loads but content missing - Increase wait time:

browser-fetch --headless --wait 10 --text URL

SSO redirect loops - Some SSO systems need manual login each time

---

Last updated: 2026-02-10