Firecrawl MCP setup: extract a page and check the result

Choose a public documentation page you can read fully in your browser. Record a section heading and a short piece of information you expect to find in the output. This small test reveals whether you received the right page, the relevant excerpt and a suitable format. You do not need to crawl an entire website to establish that.

Add the OAuth endpoint to VS Code

For browser sign-in, Firecrawl documents https://mcp.firecrawl.dev/v2/mcp-oauth. Add the following entry to .vscode/mcp.json, start the server and select the intended team during authentication.

{
  "servers": {
    "firecrawl": {
      "type": "http",
      "url": "https://mcp.firecrawl.dev/v2/mcp-oauth"
    }
  }
}

API-key access instead uses https://mcp.firecrawl.dev/v2/mcp with a bearer header. Limited keyless access is also documented on that address. The VS Code guide covers client configuration. Identify the access method before treating a different tool list as a fault.

Choose a tool for the task

Keyless access exposes Search, Scrape and Parse; authenticated availability depends on the service offering and team configuration. Connections therefore do not necessarily expose identical tools. Begin with Scrape and one page. Use that result to decide whether extending the task to Map or Crawl would help.

Check text and structured output

1. Retrieve the prepared URL and ask for readable text first. Find the section heading you recorded before the test.

2. Compare one paragraph with the browser. Include neighbouring qualifications so an exception is not lost from the extracted answer.

3. If you need structured fields, define their names and meaning. Check which values are actually supported by the same source page.

4. Repeat with a second page where one expected field is absent. A missing value should not be replaced by a similar-looking value from the navigation.

Expand the scope deliberately

Before expanding to a collection, record the starting address, allowed paths, required fields and a stopping rule. Manually inspect a small sample before retrieving more pages. For searching before extraction, use the Exa guide; for visible interaction sequences, see the Playwright guide.

Why is firecrawl_crawl missing from my tool list?

First check the endpoint and authentication. Keyless access has a smaller selection. Then compare the current vendor documentation with the team policy governing your connection.

Why is the output shorter than the browser page?

Check the output format, excerpt and exact destination URL. Short output establishes neither complete extraction nor a connection fault. Compare a specific section to identify what is missing.

Should I start a crawl for a single URL?

Begin with a single retrieval. Crawling expands the scope and makes diagnosis harder if the first expected passage is already missing.

How do I know whether a structured field is correct?

Keep the source location alongside the field value and compare it with the original page. Also test a page without that field to detect unintended substitute values.

The Firecrawl registry search provides further vendor declarations. Use the access planner to establish required functions and data scope before connecting a larger workflow.

Vendor documentation read on 3 October 2026. Run the proposed checks in your own environment.

1. Firecrawl: authentication paths
   Show retrieval command

   curl -s https://docs.firecrawl.dev/mcp-server
2. Firecrawl: tools and availability
   Show retrieval command

   curl -s https://docs.firecrawl.dev/mcp-server/tools
3. Firecrawl: keyless and API-key access
   Show retrieval command

   curl -s https://docs.firecrawl.dev/mcp-server/keyless
4. VS Code: MCP configuration
   Show retrieval command

   curl -s https://code.visualstudio.com/docs/agent-customization/mcp-servers

All posts