Pages that resist ordinary browsing
Retrieve a public page as clean Markdown or capture a truthful screenshot when JavaScript, layout, or anti-bot friction gets in the way.
scrape_as_markdown · screenshotResearch infrastructure · MCP
Decodo turns difficult public web pages into structured inputs an AI agent can actually use—then exposes the workflow through an official MCP server.
Partner link. It supports Superbash Learn at no extra cost to you.

Beginner course
Thirteen illustrated tutorials explain when scraping belongs in an agent workflow, then cover every current MCP toolset plus direct API rendering, browser actions, async jobs, batches, multi-format output, n8n, LangChain, Agent Skills, and operations.
Know when built-in search is enough and when repeatable research needs a scraping layer.
Open tutorialKnow when to use MCP, the Web Scraping API, Site Unblocker, or direct proxies.
Open tutorialSet up Basic authentication, hosted or local MCP, narrow toolsets, and a smoke test.
Open tutorialCollect Markdown or a rendered PNG with controlled geography, device, and output size.
Open tutorialUse JavaScript rendering, browser actions, waits, selectors, and XHR capture.
Open tutorialResearch Google, Bing, Ads, Lens, hotels, ChatGPT, Perplexity, and Google AI Mode.
Open tutorialCompare Amazon, Walmart, Target, and TikTok Shop without flattening price evidence.
Open tutorialCollect public community observations while keeping claims and verification separate.
Open tutorialSearch videos, inventory channels, collect metadata and subtitles, and understand downloads.
Open tutorialScale a proven job with task IDs, callbacks, batches, and multi-format results.
Open tutorialSet budgets, validate records, retry by cause, monitor cost, and enforce safe boundaries.
Open tutorialChoose visual automation, a TypeScript agent tool, or live Decodo documentation over MCP.
Open tutorialTeach coding agents when to select Decodo without putting credentials inside a Skill.
Open tutorialWhy it earns a place
Use it when your workflow needs complete page content, repeatable fields, a visual capture, or a target-specific scraper—not just a list of search snippets. The useful pattern is simple: let Decodo retrieve the evidence, let the agent interpret it, and make a human review the final result.
Best uses
Retrieve a public page as clean Markdown or capture a truthful screenshot when JavaScript, layout, or anti-bot friction gets in the way.
scrape_as_markdown · screenshotCompare products, prices, sellers, and availability across structured ecommerce targets without asking the model to guess from snippets.
Amazon · Walmart · TargetCollect current results from search engines, Reddit, YouTube, or TikTok, then make the agent cite and review what it found.
Search · Reddit · YouTube · TikTokFastest safe setup
Decodo’s current MCP documentation supports a hosted endpoint as well as its official npm server. The exact client configuration differs, but the operating discipline should not.
Create a Decodo Web Scraping API account and copy its Basic authentication token. Keep it in your client’s secret or environment configuration—never in a prompt or repository.
Use the local @decodo/mcp-server package or the hosted endpoint at https://mcp.decodo.com/mcp. Follow your agent client’s MCP configuration format.
Start with web plus one job-specific toolset such as search, ecommerce, or social_media. A smaller catalog makes tool choice easier to audit.
Name the target, required fields, failure behavior, and maximum request count. Verify the source URLs and output before turning the prompt into a repeatable workflow.
A better first prompt
Use Decodo's scrape_as_markdown tool on [PUBLIC URL].
Return: title, source URL, date, and five evidence-backed findings.
If the page fails, stop after one retry and report the error.
Cap the entire run at 2 Decodo requests.Go deeper
The video companion preserves Ron's real Amazon comparison, request-cost discussion, failures, corrections, and reusable Hermes Skill. Use the new course above for the current official MCP and API path.