Skip to main content
Glama

FineData MCP Server

MCP (Model Context Protocol) server for FineData web scraping API.

Enables AI agents (Claude, Cursor, GPT, …) to fetch pages that need a full browser and return clean data:

  • JavaScript rendering and browser actions

  • Captcha solving

  • Datacenter / ISP / residential / mobile proxies

  • Markdown or JSON output (markdown by default)

  • AI structured extraction

  • Local stdio or remote Streamable HTTP (+ OAuth 2.1)

Version: 0.3.1

Modes

Start with a plain request. Stealth and proxy modes are available when a page needs a full browser; they consume more tokens than a plain request. Exact rates are in the documentation and via get_usage. Gateway may apply a domain strategy that overrides the requested engine or proxy; trust tokens_used in the response.

Related MCP server: Thordata MCP Server

Installation

curl -LsSf https://astral.sh/uv/install.sh | sh
FINEDATA_API_KEY=fd_xxx uvx finedata-mcp

pip

pip install finedata-mcp
FINEDATA_API_KEY=fd_xxx finedata-mcp

npx

npx -y @finedata/mcp-server

Cursor

~/.cursor/mcp.json (Windows: %USERPROFILE%\.cursor\mcp.json):

{
  "mcpServers": {
    "finedata": {
      "command": "uvx",
      "args": ["finedata-mcp"],
      "env": {
        "FINEDATA_API_KEY": "fd_your_api_key_here"
      }
    }
  }
}

Cursor deeplink (after publishing): install via MCP directory / “Add to Cursor”.

Remote (Streamable HTTP)

{
  "mcpServers": {
    "finedata": {
      "url": "https://mcp.finedata.ai/mcp",
      "headers": {
        "Authorization": "Bearer fd_your_api_key_here"
      }
    }
  }
}

OAuth 2.1 (Claude.ai / ChatGPT connectors): register via AS at https://api.finedata.ai, consent in the cabinet, then use the access token as Bearer.

Environment

Variable

Required

Default

Description

FINEDATA_API_KEY

stdio: yes

—

API key

FINEDATA_API_URL

no

https://api.finedata.ai

API base

FINEDATA_TIMEOUT

no

180

HTTP client timeout (seconds)

FINEDATA_MCP_HOST

HTTP

0.0.0.0

Bind host

FINEDATA_MCP_PORT

HTTP

8080

Bind port

FINEDATA_OAUTH_ISSUER

remote OAuth

—

AS URL (gateway)

FINEDATA_MCP_RESOURCE_URL

remote OAuth

—

Public MCP URL

FINEDATA_JWT_SECRET

remote OAuth

—

Verify aud=mcp JWTs

Tools

Tool

Purpose

scrape_url

Sync scrape with GET, read-only (markdown default)

send_http_request

Sync POST / PUT / PATCH / DELETE through the same pipeline

scrape_async

Async job, GET (formats=['markdown'] default)

get_job_status

Poll job (markdown, not raw HTML)

cancel_job

Cancel job

list_jobs

List jobs

batch_scrape

Up to 100 URLs (string or {url,...} objects)

get_batch_status

Batch progress

get_usage

Period usage via api_tokens_used

Notable parameters: use_antibot, proxy_country, proxy_sticky, proxy_profile_id, auto_retry, stealth modes, formats, only_main_content, extract_*.

Async/batch: no csv/xlsx formats (sync only).

HTTP transport

finedata-mcp --transport http --host 0.0.0.0 --port 8080

Health: GET /health. Protected resource metadata is served at the path-scoped URL for the endpoint — GET /.well-known/oauth-protected-resource/mcp — because FINEDATA_MCP_RESOURCE_URL carries the /mcp path. Clients normally find it from the WWW-Authenticate header on a 401 rather than guessing.

Support

License

MIT

Available Tools

5 tools
batch_scrapeAInspect

Scrape multiple URLs in a single batch request.

Benefits:

  • Submit up to 100 URLs at once

  • Parallel processing for speed

  • Single webhook notification when all complete

Returns a batch_id and list of job_ids for tracking.

Inputs: urls (required), use_js_render, use_residential, callback_url.

get_job_statusAInspect

Get the status of an async scraping job.

Statuses:

  • pending: Job is queued

  • processing: Worker is scraping

  • completed: Success, result available

  • failed: Error occurred

  • cancelled: Job was cancelled

Poll this endpoint until status is 'completed' or 'failed'.

Inputs: job_id (required).

get_usageBInspect

Get current API usage and token statistics.

Returns:

  • Tokens used this billing period

  • Token limit for your plan

  • Usage breakdown by feature

Inputs: no named parameters.

scrape_asyncAInspect

Submit an async scraping job for long-running requests.

Use this for:

  • Pages that take > 60 seconds to load

  • Heavy JS rendering tasks

  • When you don't need immediate results

  • Batch processing workflows

Returns a job_id that you can poll with get_job_status.

Inputs: url (required), use_js_render, use_residential, use_undetected, +3 more.

scrape_urlAInspect

Scrape content from any web page with advanced antibot bypass.

Features:

  • TLS fingerprinting (Chrome, Firefox, Safari profiles)

  • JavaScript rendering for SPAs (React, Vue, Angular)

  • Captcha solving (reCAPTCHA, hCaptcha, Cloudflare Turnstile)

  • Residential and mobile proxy support

  • Automatic retry with smart detection

Use cases:

  • Extract text/HTML from any website

  • Scrape JavaScript-rendered content

  • Access pages behind Cloudflare or other protections

  • Get data from pages with captchas

Token costs:

  • Base request: 1 token

  • Antibot bypass: +2 tokens

  • JS rendering: +5 tokens

  • Residential proxy: +3 tokens

  • Captcha solving: +10 tokens

Inputs: url (required), use_js_render, use_residential, use_undetected, +4 more.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A4.1/5.0

    Scored across 5 tools

    Disambiguation5/5

    Each tool has a clearly distinct purpose with no ambiguity: batch_scrape handles multiple URLs in parallel, scrape_async is for long-running single jobs, scrape_url is for immediate single requests with advanced features, get_job_status tracks async/batch jobs, and get_usage monitors API consumption. The descriptions explicitly differentiate use cases and workflows.

    Naming Consistency4/5

    The naming follows a consistent snake_case pattern with clear verb_noun structures (e.g., batch_scrape, get_job_status, scrape_async). However, scrape_url deviates slightly by using a noun_verb format instead of a verb_noun pattern, which is a minor inconsistency in an otherwise predictable set.

    Tool Count5/5

    With 5 tools, this server is well-scoped for web scraping operations. Each tool earns its place by covering distinct aspects of the workflow: submission (batch, async, immediate), status tracking, and usage monitoring. This count is neither too sparse nor bloated for the domain.

    Completeness5/5

    The tool surface provides complete coverage for the web scraping domain. It supports all key operations: multiple submission methods (batch, async, immediate), job lifecycle management (status tracking), and administrative functions (usage monitoring). There are no obvious gaps that would hinder agent workflows.

    Related MCP Connectors

    Related MCP Servers