Using n8n to connect to firecrawl-mcp-server throws a 404 error

[wuhuanyan]

1. Self-hosted Firecrawl Service Test

I have self-hosted the firecrawl service and conducted some basic tests using cURL.

curl -X POST http://127.0.0.1:3002/v1/scrape \
    -H 'Content-Type: application/json' \
    -d '{
      "url": "https://www.baidu.com",
      "formats": ["markdown", "html"]
    }'

✅ This test works fine — I can successfully retrieve content from Baidu in both markdown and HTML formats.

---

2. Self-hosted firecrawl-mcp-server with SSE_LOCAL enabled

Then, I also self-hosted the firecrawl-mcp-server using Docker Compose, and set SSE_LOCAL: true.

services:
  firecrawl:
    image: mcp/firecrawl
    container_name: firecrawl-mcp-server
    environment:
      FIRECRAWL_API_URL: "http://host.docker.internal:3002/v1"
      FIRECRAWL_RETRY_MAX_ATTEMPTS: "5"
      FIRECRAWL_RETRY_INITIAL_DELAY: "2000"
      FIRECRAWL_RETRY_MAX_DELAY: "30000"
      FIRECRAWL_RETRY_BACKOFF_FACTOR: "3"
      FIRECRAWL_CREDIT_WARNING_THRESHOLD: "2000"
      FIRECRAWL_CREDIT_CRITICAL_THRESHOLD: "500"
      SSE_LOCAL: true
    stdin_open: true
    tty: false
    restart: "no"
    ports:
      - "3000:3000"
    extra_hosts:
      - "host.docker.internal:host-gateway"
    networks:
      - default

networks:
  default:
    driver: bridge

✅ The server starts up normally, and I can access it via port 3000.

---

3. Integration with n8n

In n8n, I managed to establish communication with the firecrawl-mcp-server, as evidenced by its ability to fetch the tool list.

❌ However, when running the MCP client test in n8n, I receive a 404 Not Found error.

There are no logs generated from either:

• firecrawl (the scraping backend)
• firecrawl-mcp-server (the MCP adapter)

Since a 404 is returned, I suspect the request actually reaches the firecrawl-mcp-server. Is it possible that detailed logging is not enabled by default in firecrawl-mcp-server?

Without proper logs, debugging this issue becomes very difficult.

---

Summary of Questions

1. Why does the MCP client test return a 404 error?
2. Why are there no logs being generated on either service?
3. Is there a way to enable more verbose logging in firecrawl-mcp-server for debugging purposes?

Any help or suggestions would be greatly appreciated!

[wuhuanyan]

I turned off firecrawl for testing purposes.
n8n shows an error saying it can't connect to the firecrawl service.
This suggests that the firecrawl-mcp-server is working.
However, its request URL doesn't match that of firecrawl — could this be causing the error?

[ErixWong]

I encountered a similar issue while using Cherry Studio. Both my Firecrawl and Firecrawl MCP servers are hosted locally in Docker containers. Postman tests confirmed Firecrawl works properly. The MCP server successfully generated tools after I entered its URL, but when the LLM attempted to invoke it, I received a 404 error with no logs.