Browserless MCP in LM Studio
LM Studio gives local models Browserless's browser tools by connecting to the hosted MCP server from its mcp.json file. Once it's connected, LM Studio can scrape, search, crawl, and drive a cloud browser with the Browserless MCP tools.
- A Browserless API token from your account dashboard, or a Browserless account for OAuth sign-in
- LM Studio 0.3.17 or later
- A downloaded model that supports tool calling
Connect to the hosted server
The hosted server runs at https://mcp.browserless.io/mcp, so there's nothing to install.
In LM Studio 0.3.17 or later, open the Program tab in the right sidebar, click Install, then Edit mcp.json. Add:
{
"mcpServers": {
"browserless": {
"url": "https://mcp.browserless.io/mcp",
"headers": {
"Authorization": "Bearer YOUR_API_TOKEN_HERE"
}
}
}
}
On LM Studio 0.4.10 or later, you can sign in with OAuth instead: drop the headers block and LM Studio opens a browser window to approve access. Earlier versions need the API token header.
LM Studio asks you to confirm each tool call, and you can allow a tool once or always. Change those choices later under App Settings > Tools & Integrations.
To use a region other than US West, add an x-browserless-api-url header, or a browserlessUrl query parameter on the server URL for clients that only take a URL. See regional endpoints.
Run the server locally
To run the MCP server process on your machine instead, use the @browserless.io/mcp npm package over stdio. You need Node.js 24 or newer. Browser work still runs on Browserless, so this is mainly for pinning a server version (add it to the package name, for example @browserless.io/mcp@1.41.0) or pointing at a self-hosted instance with BROWSERLESS_API_URL. Both setups use the server name browserless, so replace the hosted entry with this one rather than adding both.
In the same mcp.json:
{
"mcpServers": {
"browserless": {
"command": "npx",
"args": ["-y", "@browserless.io/mcp"],
"env": {
"BROWSERLESS_TOKEN": "YOUR_API_TOKEN_HERE"
}
}
}
}
Check the connection
Save mcp.json. LM Studio loads the server automatically. Start a chat with a model that supports tool calling and ask it to use a Browserless tool. LM Studio asks you to confirm the first call.
FAQ & Troubleshooting
I can't find mcp.json on disk
Its location has varied between LM Studio versions and platforms. Always edit it through Program → Install → Edit mcp.json in the app.
The model never calls the Browserless tools
The model needs tool-calling support. Pick a model marked as supporting tool use in LM Studio, and name the tool in your prompt the first time, for example "use browserless_smartscraper to read this page".
The server connects but every tool call returns 401 Unauthorized
The token is missing or wrong. Copy it again from your account dashboard and check the header reads Bearer followed by the token, with no quotes or trailing whitespace inside the value. If you used an environment variable, confirm it's set in the environment the client was launched from. GUI apps started from the dock or Start menu don't see variables exported in your shell profile.
The connection fails with an SSE error like Non-200 status code (400)
The hosted server speaks Streamable HTTP only and doesn't support the legacy SSE transport. Check the client entry is using its HTTP transport (see the config above), not an SSE one.