M
MCP Directory CURATED DIRECTORY
911 MCP servers indexed
中文
← Back to directory
D

Decodo MCP Server

Official
Enable your AI agents to scrape and parse web content dynamically, including geo-restricted sites
How to install ↓ Compare GitHub source repository ↗
Category
Dev Tools #158 of 438
Stars
★ 37 Popular
Transport
stdio (local process)
Runtime
Node.js 18+ · Hosted (no local runtime)
Credentials
API key / credential required
Last commit
Tools
30
57FMRS · C

An officially maintained MCP server that wraps Decodo's commercial Web Scraping API into 30 tools spanning SERP, eCommerce, social media, and AI search, with modular toolsets and geo controls; well suited for production-grade real-time web data in AI agents, though it depends on a third-party cloud service and token auth.

Strongest · Documentation 16/20 Weakest · Reliability 7/20

Reliability
7/20
Security and permissions
12/20
Maintenance
9/20
Documentation
16/20
Setup experience
13/20
Why each score
Reliability 7/20
The only materials available are the server.json manifest and README—no CI workflow files or committed tests are present to verify that the MCP init/tool-list handshake actually works as described. The README enumerates 25+ tools (scrape_as_markdown, google_search, amazon_*, etc.), but the manifest only declares a single npm package with stdio transport, leaving no static evidence that the described tool behaviors match the real implementation, nor any visibility into error handling or dependency-failure controllability. Per the static calibration rule, without verifiable execution evidence (real CI plus committed tests) the score cannot exceed 12, so this is scored as plausible-but-unverified.
Security and permissions 12/20
Credentials are passed via the SCRAPER_API_TOKEN env var or an Authorization: Basic header; all README examples use placeholders (e.g. <basic_auth_token>) and no real secrets are exposed, so no red line is triggered. The server's core purpose—scraping arbitrary sites, bypassing geo-restrictions, and querying e-commerce/social targets via a large residential/mobile proxy pool (125M+ IPs, 195+ locations)—means substantial outbound traffic is routed through Decodo's third-party infrastructure; this is disclosed at a high level but lacks detail on data retention, logging, or third-party handling policies. The toolsets mechanism is a positive least-privilege feature (tools can be scoped down), but by default, with no toolsets specified, all tools are registered, so least privilege is not the default. Operations are largely read-only (scraping/screenshots) with no irreversible destructive defaults observed, avoiding red-line status, but scoping and data-flow disclosure remain incomplete.
Maintenance 9/20
The repository-level license field is reported as unknown, while the README claims an MIT license and links to the LICENSE file of a different repository (Decodo/Decodo) rather than this repo's own license file—an inconsistency between documentation and repo metadata that signals incomplete governance. The current version (1.2.3) suggests some iterative release cadence, and with 34 stars, 3 open issues, and no archived status, the project appears actively maintained rather than abandoned. However, there is no visible release history, changelog, security-response channel, or further evidence of maintainer identity/ownership to judge issue-response speed or long-term maintenance commitment.
Documentation 16/20
The README is thorough, covering an overview, quick start, concrete configuration examples for Claude Desktop, Cursor, and Windsurf, local build/run instructions, a toolset breakdown table, a full tool list with example prompts, a parameter reference table, and usage examples for geo-restriction bypass and token-limit truncation—well-layered with sufficient examples. Gaps remain: no per-tool cost/quota detail beyond a general '2K free requests' mention, no systematic troubleshooting section for API errors, rate limits, or auth failures, and the license claim contradicts the repo's own metadata, constituting a hidden assumption—so it falls short of full marks.
Setup experience 13/20
The docs provide concrete configuration examples for both hosted mode (URL + Basic Auth) and local mode (npx or manual clone/build), covering the three mainstream clients (Claude Desktop, Cursor, Windsurf) with clear step sequences. However, the local path requires manually cloning, running npm install/npm run build, and hand-assembling an absolute path into the client config—a manual, fragile installation flow. Combined with the static calibration rule capping setup at 15 absent CI/test evidence, this is scored as clear-but-manual rather than fully polished.

Static review · not runListed 2026-09-07

Read the FMRS scoring method →

Fit and risk

What it can accessUses the network

Best for

  • Developers who need real-time web data for LLMs and RAG pipelines
  • Teams building eCommerce intelligence, social monitoring, or price aggregation tools
  • AI agent projects that want production scraping without maintaining proxy infrastructure

Not for

  • Users who cannot rely on a third-party paid scraping service or hosted endpoint
  • Workflows needing only local file or internal data access
  • Privacy-sensitive cases where sending target URLs and prompts to Decodo's servers is unacceptable

Required permissions

  • Requires a Decodo Web Scraping API basic auth token (SCRAPER_API_TOKEN)
  • Downloads and runs the @decodo/mcp-server package via npx from npm
  • Requires network access to call Decodo's hosted scraping service

Risks and side effects

  • All target URLs and prompts pass through a third-party (Decodo) server
  • API usage costs apply once the free quota is exhausted
  • Scraping target sites may violate their terms of service; verify compliance yourself
  • Dependency on an external npm package and hosted API introduces supply-chain and availability risks

Setup

Before you start

Runtime：Node.js 18+ · Hosted (no local runtime)

SCRAPER_API_TOKEN requiredsecret Basic auth token for the Web Scraping API, obtained from the dashboard at dashboard.decodo.com.
Other optional settings (1)
TOOLSETS optional Optional comma-separated toolset list (web, search, ecommerce, social_media, ai) to register only selected tools; if unset, all tools are registered.
  1. Create a free account at dashboard.decodo.com (up to 2K free requests) and get a Web Scraping API basic auth token.
  2. Install Node.js 18+.
  3. In Claude Desktop's claude_desktop_config., add a server with command npx, args ["-y","@decodo/mcp-server"], env SCRAPER_API_TOKEN set to the token, and optional TOOLSETS such as web,ai.
  4. Save and restart the client. Alternatively, use the hosted endpoint https://mcp.decodo.com/mcp with a Basic Authorization header.
claude_desktop_config.json 
{"mcpServers":{"Decodo":{"command":"npx","args":["-y","@decodo/mcp-server"],"env":{"SCRAPER_API_TOKEN":"<basic_auth_token>","TOOLSETS":"web,ai"}}}}

Shown for Claude Desktop. Other clients may use a different file or key (VS Code uses "servers") — the configurator below converts it.

.vscode/mcp.json
{
  "servers": {
    "Decodo": {
      "command": "npx",
      "args": [
        "-y",
        "@decodo/mcp-server"
      ],
      "env": {
        "SCRAPER_API_TOKEN": "<basic_auth_token>",
        "TOOLSETS": "web,ai"
      }
    }
  }
}

Goes in your project's .vscode/mcp.json (VS Code uses a "servers" key).

Terminal
claude mcp add Decodo -e 'SCRAPER_API_TOKEN=<basic_auth_token>' -e TOOLSETS=web,ai -- npx -y @decodo/mcp-server

Run it in a terminal; replace any <…> placeholders with your own values first.

Generate and validate client config

Check that it works

After configuring, prompt "Scrape the titles of the top 5 articles from Hacker News" — a structured list should return within seconds; an auth error means the token should be re-checked from the dashboard.

Troubleshooting

  1. On auth errors, double-check the token from the dashboard is configured as Basic auth
  2. Verify Node.js 18 or newer is installed
  3. Confirm the MCP client config is saved and the client restarted
  4. If context overflows, ask for fewer tokens or full content in the prompt
  5. For geo-restricted targets, use the geo parameter to change the exit IP country

Things to try

Once connected, you can ask your AI assistant things like:

  • Scrape peacock.com from a US IP address and tell me the pricing.
  • Scrape Google Search for wireless keyboards and tell me the top result.
  • Get the metadata for YouTube video dFu9aKJoqGg.
  • Ask Perplexity what the latest trends in web development are.

Tools 30

scrape_as_markdown read-only
Scrapes any target URL and returns results in Markdown.
screenshot read-only
Captures a screenshot of any webpage and returns it as a PNG image.
google_search read-only
Scrapes Google Search for a given query and returns parsed results.
google_ads read-only
Scrapes Google Ads search results.
google_lens read-only
Scrapes Google Lens image search results.
google_ai_mode read-only
Scrapes Google AI Mode (Search with AI) results.
google_travel_hotels read-only
Scrapes Google Travel Hotels search results.
amazon_search read-only
Scrapes Amazon Search for a given query and returns parsed results.
Show 22 more tools
amazon_product read-only
Scrapes an Amazon product page.
amazon_pricing read-only
Scrapes Amazon product pricing information.
amazon_sellers read-only
Scrapes Amazon seller information.
amazon_bestsellers read-only
Scrapes the Amazon Bestsellers list.
walmart_search read-only
Scrapes Walmart Search for a given query and returns parsed results.
walmart_product read-only
Scrapes a Walmart product page.
target_search read-only
Scrapes Target Search for a given query and returns parsed results.
target_product read-only
Scrapes a Target product page.
tiktok_post read-only
Scrapes a TikTok post URL for structured data (engagement, caption, hashtags).
tiktok_shop_search read-only
Scrapes TikTok Shop Search for a given query and returns parsed results.
tiktok_shop_product read-only
Scrapes a TikTok Shop product page.
tiktok_shop_url read-only
Scrapes a TikTok Shop page by URL.
youtube_metadata read-only
Scrapes YouTube video metadata.
youtube_channel read-only
Scrapes YouTube channel videos.
youtube_subtitles read-only
Scrapes YouTube video subtitles.
youtube_search read-only
Searches YouTube videos.
reddit_post read-only
Scrapes a specific Reddit post.
reddit_subreddit read-only
Scrapes Reddit subreddit results.
reddit_user read-only
Scrapes a Reddit user profile and their posts/comments.
bing_search read-only
Scrapes Bing Search results.
chatgpt read-only
Search and interact with ChatGPT for AI-powered responses.
perplexity read-only
Search and interact with Perplexity for AI-powered responses.

Use cases

AI-powered web scraping with fresh data instead of static training data
RAG pipelines pulling live Google, Bing, and AI search results
eCommerce intelligence: track prices, listings, and sellers across marketplaces
Social media data collection from Reddit, TikTok, and YouTube
Travel and price aggregation tools
Accessing geo-restricted content

Supported clients

Claude Desktop
Cursor
Windsurf
Claude CodePartial support

Listed from the project's documentation, not tested by this site.

Overview

The Decodo MCP Server is a web scraping layer for AI agents that connects MCP-compatible clients to Decodo's Web Scraping API. It scrapes websites, search engines, eCommerce platforms, and social media, returning structured outputs in Markdown, JSON, and screenshots, with server-side JavaScript rendering and anti-bot handling backed by 125M+ residential IPs across 195+ locations. Tools are organized into modular toolsets (web, search, ecommerce, social_media, ai) that can be selectively enabled.

Similar servers

Context7 80 · B

Upstash's official server providing up-to-date third-party library docs for AI coding assistants

★ 63.3k · Tools 2 Compare with this →
Compare with other MCP servers

Source revision 62b90162dcf4 Data synced 2026-10-19 Read the FMRS scoring method