Choose the right Decodo scraping route

Decodo is not one scraper. It offers several ways to reach public web data. Beginners usually lose time by starting too low in the stack: they configure proxies when an MCP tool could already complete the job.

Use this rule:

• Talking to an AI assistant: start with the Decodo MCP server.
• Building an app, scheduled job, or data pipeline: use the Web Scraping API.
• Maintaining your own browser or scraper: consider Site Unblocker or a proxy product.

This course concentrates on MCP and the Web Scraping API because they remove most proxy rotation, rendering, parsing, and retry work.

What each route is for

Route	Best first use	What you control
MCP server	Ask Claude, Cursor, Windsurf, Hermes, or another MCP client to collect current evidence	Toolsets, prompt, geography, rendering, device, output limit
Web Scraping API	Put scraping inside code or automation	Endpoint, target template, request mode, browser actions, response formats
Site Unblocker	Send traffic from your own scraper through managed anti-bot handling	Proxy-style connection, location, session, headers, rendering
Residential/mobile/ISP/datacenter proxies	Run a custom network stack with direct IP control	Pool, location, rotation, sticky sessions, protocol

Do not begin with proxies because they sound more powerful. Begin with the output you need.

MCP toolsets in plain English

Decodo groups its MCP tools into five toolsets. Enabling only relevant toolsets keeps the assistant’s choices clear.

1. web — any public URL as Markdown, or a PNG screenshot.
2. search — Google Search, Google Ads, Google Lens, Google Travel Hotels, and Bing Search.
3. ecommerce — Amazon, Walmart, Target, and TikTok Shop search/product tools.
4. social_media — Reddit posts, subreddits and users; TikTok posts; YouTube search, channels, metadata, and subtitles.
5. ai — ChatGPT, Perplexity, and Google AI Mode results.

All 30 current MCP tools appear across later tutorials. You do not need all 30 enabled for every conversation.

Step-by-step: define one useful job

Before opening the dashboard, write four lines:

1. Target: one public URL, query, product ID, channel, or post.
2. Deliverable: exact fields or decision you need.
3. Freshness: when the information must have been observed.
4. Budget: maximum Decodo requests and retry count.

Example:

Target: the five leading pages for “best compact espresso machine”.
Deliverable: rank, title, URL, domain, snippet, and observed date.
Freshness: collect today.
Budget: 1 Google Search request; no automatic retries.

That is a scraping job. “Research espresso machines” is only a topic.

Choose a route

Use this quick test:

• Need one answer inside a chat? MCP.
• Need the same job every day? Prove it with MCP, then move the stable request to the API.
• Need clicks, typed inputs, waits, or XHR capture? Web Scraping API browser actions.
• Need raw network control for custom code? Site Unblocker or proxies.

What success looks like

A successful first scrape includes:

• the exact Decodo tool or API target used;
• source URL or query;
• observation time and requested geography;
• structured fields, not a loose paragraph;
• honest empty fields instead of invented values;
• request count and any retries;
• a human check against the source.

Safety boundary

Keep this course to public or authorized data. Do not use it for post-login content, private user information, bypassing access controls, or collecting more personal data than the task needs. Website terms, privacy rules, and local law still apply.

Next: connect one MCP client and prove that it can see Decodo’s tools.

Official references

• Decodo Web Scraping API introduction
• Decodo MCP server
• Decodo Help Documentation