Skip to directory

About Browserless MCP

Browserless provides a hosted MCP server for browser automation, scraping, search, crawling, page interaction, exports, audits, and account usage.

What you can use it for

  • Operate a remote browser for multi-step interactions
  • Scrape, search, crawl, audit and export web pages
  • Generate screenshots/PDFs and inspect Browserless account usage
Connection and setup
Hosting
Remote hosted; local npm client process can call Browserless cloud or a self-hosted Browserless instance
Transport
Streamable HTTP
Endpoint
https://mcp.browserless.io/mcp
Authentication
OAuth sign-in or API token via Authorization header or token query parameter.

Requirements

  • Browserless account for OAuth or API token access
  • MCP-compatible client
  • Node.js 24+ only if using the local npm package; hosted endpoint requires no local installation

Setup steps

  1. Connect the hosted Streamable HTTP endpoint https://mcp.browserless.io/mcp in the client, or use the official local package @browserless.io/mcp with Node.js 24+.
  2. Use the client OAuth flow to sign in, or store the API token in a secure client secret/environment variable and send it as Authorization: Bearer <BROWSERLESS_API_TOKEN>.
  3. Use a query token only when the client cannot send authorization headers and its configuration requires a URL; protect the config accordingly.
  4. Reconnect/reload the client and verify by listing Browserless tools. The Browser Agent supports one-shot sessions by default; multi-turn sessions require keepSessionAlive: true on the first call and sessionId on subsequent calls.
Tools 14

The current official docs enumerate 14 tools across browser-agent, REST API, and read-only account groups. Tool calls consume Browserless account units.

  • browserless_agent

    Drive a stateful browser across multiple steps.

    01
  • browserless_smartscraper

    Scrape pages with cascading extraction strategies.

    02
  • browserless_function

    Run custom Puppeteer JavaScript.

    03
  • browserless_export

    Export a page as HTML, PDF, or image.

    04
  • browserless_search

    Search web, news, or images.

    05
  • browserless_map

    Discover URLs on a site.

    06
  • browserless_performance

    Run a Lighthouse audit.

    07
  • browserless_crawl

    Crawl and scrape pages from a site.

    08
  • browserless_skill

    Load an agent skill or site recipe.

    09
  • browserless_account

    Read account plan and unit balance.

    10
  • browserless_usage

    Read account request and unit usage.

    11
  • browserless_logs

    Read request logs.

    12
  • browserless_sessions

    List browser sessions.

    13
  • browserless_profiles

    List saved authentication profiles.

    14
Limitations and notes
  • Local npm installation runs the MCP process locally but browser execution still uses Browserless cloud unless configured for a self-hosted instance.
  • An API token is required unless connecting through OAuth; account/plan limits apply.
  • MCP server v1.33.0+ sessions are one-shot by default; callers must opt into persistent multi-step sessions.
  • Claude.ai custom connectors require putting a token in the URL query when using the documented URL-only form.
Original sources
  • official tool reference

    Current official docs enumerate 14 tools across browser agent, REST API, and account tool groups.

  • official documentation

    Current official docs enumerate 14 tools across browser agent, REST API, and account tool groups.

Documentation checked 2026-10-28T12:35:00.000Z. Connection has not been runtime tested.

Similar tools