Browserless MCP Server – Browser Automation for AI
Browserless provides hosted and local MCP access to its browser automation platform. AI agents can drive stateful browser sessions, scrape and crawl sites, run Puppeteer code, export pages, download files, and run Lighthouse-style performance checks without managing browser infrastructure.
Overview
Browserless MCP Server connects MCP-compatible AI assistants to Browserless, a managed browser automation platform. It is designed for agents that need to interact with the live web rather than only reason over static text. Browserless documents a hosted MCP endpoint as well as a local npm-based server, so teams can choose between a zero-install remote connection and a locally launched process that can be pinned or pointed at a self-hosted Browserless endpoint.
What the MCP server enables
The server exposes Browserless browser automation tools to the AI client. Its stateful browserless_agent tool can keep a browser session alive across turns so an assistant can navigate pages, inspect snapshots, click, type, fill forms, and continue a multi-step flow in the same browser context. It also provides stateless REST-backed tools for common tasks: smart scraping, running custom Puppeteer JavaScript, triggering and retrieving downloads, exporting pages as HTML/PDF/images/offline archives, searching the web with optional scraping of results, mapping site URLs, running Lighthouse audits, and crawling sites from a seed URL.
When to use it
Use this MCP server when an AI workflow needs a real browser or Browserless REST capabilities: extracting content from pages that require JavaScript, crawling a public site, exporting a page to PDF, collecting screenshots or downloadable files, checking page performance, or automating a multi-step login and form-filling flow. It is also useful when you want browser execution handled by Browserless cloud infrastructure or by a self-hosted Browserless deployment rather than running browsers directly on a user's workstation.
Connection and authentication
Browserless provides the hosted MCP endpoint https://mcp.browserless.io/mcp. It supports OAuth for clients that can complete Browserless account login, and API-token authentication through the recommended Authorization: Bearer ... header or a token query parameter for URL-only clients. The local server is published as @browserless.io/mcp and starts in stdio mode by default with BROWSERLESS_TOKEN. It can also run a local HTTP Stream endpoint by setting TRANSPORT=httpStream and PORT.
Key considerations
A Browserless account and token are required for token-based use. Running the local MCP server does not automatically run a local browser; by default it still calls Browserless cloud for browser execution unless BROWSERLESS_API_URL points to a self-hosted instance. The default Browserless API region is US West, with documented London and Amsterdam alternatives available through x-browserless-api-url, browserlessUrl, or BROWSERLESS_API_URL. Treat browser automation as an active capability: review prompts and permissions carefully before allowing an agent to interact with authenticated sessions, forms, downloads, or custom Puppeteer code.
Supported Transports
streamable_http
URL: https://mcp.browserless.io/mcp
stdio
Command: npx
Args:
-y@browserless.io/mcp
streamable_http
URL: http://localhost:8080/mcp
Frequently Asked Questions
- When should an AI agent use Browserless MCP Server?
- Use it when the agent needs live browser automation: scraping JavaScript-heavy pages, crawling a site, generating screenshots or PDFs, downloading browser-created files, running performance audits, or completing multi-step workflows in a persistent browser session.
- What does Browserless MCP Server add to an AI agent's capabilities?
- It gives the agent access to Browserless-hosted browser execution through MCP tools, including a stateful browser agent plus stateless tools for smart scraping, custom Puppeteer functions, downloads, page exports, search, site mapping, performance audits, and crawling.
- What can the AI agent access or manage through this server?
- The agent can operate browser sessions and Browserless REST-backed tasks according to the connected Browserless account and token. It can retrieve page content, exported files, search results, crawl output, and audit results, but it should only be granted access to sites and sessions the user is authorized to automate.
- How is authentication configured for Browserless MCP Server?
- The hosted endpoint supports OAuth for compatible clients and API-token authentication by `Authorization: Bearer ${BROWSERLESS_TOKEN}` or a `token` query parameter for URL-only clients. The local npm server expects `BROWSERLESS_TOKEN`; optional settings include `BROWSERLESS_API_URL`, `TRANSPORT`, `PORT`, `BROWSERLESS_TIMEOUT`, `BROWSERLESS_MAX_RETRIES`, and `BROWSERLESS_CACHE_TTL`.
- Which transport should be used for Browserless MCP Server?
- Use the hosted Streamable HTTP endpoint when you want no installation and can authenticate with OAuth or a Browserless token. Use stdio with `npx -y @browserless.io/mcp` for local desktop MCP clients. Use the local HTTP Stream endpoint only when you intentionally run the npm server as a long-lived shared endpoint with `TRANSPORT=httpStream`.