Skip to main content
Add Firecrawl’s Search, Scrape, and Crawl tools to LM Studio through the hosted MCP server, so your local models can read the live web. LM Studio 0.3.17 and later act as an MCP host and read servers from an mcp.json file in the same notation Cursor uses.

​
Quick setup

  1. Create a Firecrawl API key.
  2. In LM Studio, open the Program tab in the right sidebar and choose Install > Edit mcp.json.
  3. Add Firecrawl under mcpServers and save:
If mcp.json already lists other servers, copy only the "firecrawl": { ... } entry into the existing mcpServers object. LM Studio loads the server when you save the file. Open a chat with a model that supports tool use and confirm that firecrawl_search and firecrawl_scrape appear in the chat’s tool list. Keep the API key out of shared screenshots and source control. LM Studio keeps mcp.json in its data folder in your home directory (~/.lmstudio/mcp.json).

​
Sign in instead of using a key

LM Studio supports MCP OAuth for remote servers. Add https://mcp.firecrawl.dev/v2/mcp-oauth as the server url without headers, then follow the browser sign-in. See For Humans.

​
Run the local server instead

To run Firecrawl on your own machine, add a command entry that starts the npm package instead of the url entry. See the local MCP server guide for the current command and options.

​
Try it

​
Troubleshooting

  • Tools do not appear: make sure the model supports tool use, then check that mcp.json is valid JSON and that the firecrawl entry sits inside mcpServers.
  • Authentication fails: confirm the header reads Bearer followed by your key with a single space, and that the key is active in your Firecrawl dashboard.
  • The model slows down or runs out of context: Firecrawl tools can return long pages. Ask for specific URLs, use a model with a larger context window, or enable only the tools you need.
  • Review tool calls: LM Studio asks for confirmation before running MCP tools by default. Only install MCP servers from sources you trust.