Zyte’s blog on September 30, 2026 introduces Zyte MCP Server—a hosted Model Context Protocol server that makes Zyte API and Scrapy Cloud available to MCP-compatible AI and coding agents (blog, Valter Sciarrillo). Companion docs cover connect, tools, auth, and billing.
What it is
This is not a new extraction engine. It is the same Zyte API and Scrapy Cloud surface, reached through a remote Streamable HTTP endpoint at https://mcp.zyte.com/v1/mcp, with no server package to maintain on the client side. Zyte names Claude Code, Cursor, Codex, GitHub Copilot, and VS Code (docs also list claude.ai) as MCP clients that can connect.
Product-level capabilities include fetching protected or JavaScript-heavy pages; returning clean Markdown (with browser rendering, screenshots, and geolocation where needed); web search with structured results; structured field extraction; and natural-language queries over usage, spend, and success rates.
Auth: OAuth in the agent config
On first connection the client completes a Zyte sign-in. The blog’s framing: OAuth gives the client a short-lived, scoped token instead of placing a permanent Zyte API key in the agent’s MCP configuration. Docs also describe an organization mcp_access_key used for metering and limits after sign-in—so the lock is “no permanent Zyte API key pasted into agent MCP config,” not “no keys exist anywhere in the org.”
Scrapy Cloud from the agent
Beyond one-shot fetch, Zyte MCP can start, stop, and monitor Scrapy Cloud jobs from deployed spiders; inspect job state, logs, stats, and item samples; and create and manage recurring schedules. Docs warn that agents can cancel jobs, change settings, and write or delete collections and schedules (some deletes are not undoable)—treat those as destructive ops subject to the user’s approval controls in the client.
Pricing shape
Zyte says there is no separate MCP fee: MCP use is free at the protocol layer, and usage is metered through existing Zyte API and Scrapy Cloud pricing and rate limits. Docs mention a trial credit for new users; we do not treat that as permanent pricing.
Marketing claims to attribute
Zyte cites an internal Markdown cleanup benchmark (35 HTML cleanup / HTML-to-Markdown tools plus three hosted services across 299 pages) with high main-content retention and token reduction—vendor figures, not an ATN bench. Lines about “#1” unblocking, 320,000 access strategies, and highest success rate stay Zyte-attributed marketing, not rankings we verify.
Who this is not
This is hosted Zyte API + Scrapy Cloud over MCP—not a Firecrawl scrape MCP lead, not Bright Data’s MCP story, not a local Scrapy MCP package as the primary path, and not a generic browser MCP.
Who should care
Teams already on Zyte who want coding agents to fetch clean Markdown and operate Scrapy Cloud without embedding a permanent API key in MCP config should start at the announcement and MCP docs. Confirm OAuth + metering steps and approval settings for Scrapy job control before rolling it into shared agent setups.
The Campfire
No commentsNobody has pulled up a log by this one yet. Be the first to say what you make of it.
Held for the desk. It appears after a look.