sse server There is an error in the return parameter type.

[lijiaxing1997]

see server return
ToolParameter(name='location', parameter_type='object', description='Location settings for scraping', required=False, default='{}', parameter_items=None)], metadata={'endpoint': 'http://localhost:3000/sse'}, identifier='firecrawl_scrape')
When using it, there will be an error.
code": "invalid_type",\n "expected": "string",\n "received": "object",\n "path": [\n "location"\n ],\n "message": "Expected string, received object"\n }\n]', 'annotations': None}

[dabdine]

This appears to be easily reproducible using the mcp inspector:

 npx @modelcontextprotocol/inspector

Launch the local firecrawl mcp server:

env SSE_LOCAL=true FIRECRAWL_API_KEY=xxx npx -y firecrawl-mcp

Now list tools in the inspect -> execute the firecrawl_search tool using:

• query: test
• country: {"country":"us"}

Using mcp python v1.9.3.

Mostly problematic when using OpenAI in strict mode, where you're specifying all the parameters as required. OpenAI will generate a tool call with a location that meets the spec, but the spec will fail when it gets to the server due to this issue.

This appears to be due to misalignment between the firecrawl mcp server's search tool and the firecrawl search API (https://docs.firecrawl.dev/api-reference/endpoint/search). The search API location parmater is a string, but in the MCP server it's structured content (an object): https://github.com/mendableai/firecrawl-mcp-server/blob/main/src/index.ts#L462. The python client itself also requires a string for the location: https://github.com/mendableai/firecrawl/blob/main/apps/python-sdk/firecrawl/firecrawl.py#L588.