- SPECIMEN
Zyte MCP
- FILED AS
Remote MCP server that gives coding agents Zyte API scraping and Scrapy Cloud access
- INTAKE DATE
- 2026-10-07
- CLASS
- Coding
- METHOD
- 2 pages scraped, 3 search passes, 3 community sources
Fetching, extraction, search and Scrapy Cloud project work all come from one server, though no public benchmark yet shows how its extraction quality compares with rivals. THE RUNDOWN
Each documented client needs a single add command plus a login, and Zyte ships a smoke test that confirms the call went through a Zyte MCP tool. ZERO TO RUNNING
Billing only successful responses and discounting commitment plans 25%-52% earns credit, while a $5 trial leaves little room to test a real crawl. THE DAMAGE
Zyte's own pages carry no dated star count, release date or user figure for this server, and public discussion is limited to directory listings on Glama and mcpservers.org. THE STREET
Koda Score = weighted blend: capability 35, ease 20, value 25, momentum 20.
Zyte MCP hands an agent page fetching, structured extraction, web search and Scrapy Cloud control through one remote endpoint, billed to a dedicated mcp_access_key you can track and limit. The catch is the meter: Zyte's pricing page says request cost depends on the target website and selected features, so a crawl's price stays unknown until you run the estimator or a small test. Scrapy users and builders who want agent-driven extraction under a spending ceiling should adopt it now.
Zyte's docs show one add command and one login putting its scraping API and your Scrapy Cloud projects inside a coding agent, with every request billed to a key you can cap.
Captured: docs.zyte.com/zyte-web-data/mcp.html on 2026-10-07.
The short version.
3 LINESName Zyte in your prompt, since the docs warn an agent may otherwise reach for its own web tools.
Cap spend on the dedicated mcp_access_key before you let an agent loose on a crawl.
Start with the $5 one-month trial credit Zyte's pricing page lists, and set a spending limit before it ends.
What it actually does.
6 CAPABILITIES- Page fetching
- Your agent retrieves web pages through Zyte API on your behalf.
- Structured extraction
- Pulls data out of fetched pages, and the Glama listing notes it routes among extraction types such as detail and list pages.
- Web search
- Lets the agent run web searches through Zyte.
- Scrapy Cloud projects
- Works with your Scrapy Cloud projects so crawl jobs can be managed from the editor.
- Dedicated spend key
- Agent requests run on an organization key you can monitor and restrict separately from your other usage.
- Organization picker
- If you belong to several Zyte organizations, the agent asks which one to use.
How you would actually use it.
2 PLAYS- Single-page extraction test
- Point the agent at one product listing page you already copy by hand and have Zyte return the items as JSON; diff the output against your manual copy.WHODeveloper evaluating scraping vendorsPAYOFFA quick read on extraction accuracy and real per-request cost before any crawl runs.
- Agent-run Scrapy Cloud chores
- Keep deploying spiders as usual and let the coding agent check job status and start runs through Zyte MCP while you edit spider code.WHOTeams running Scrapy Cloud spidersPAYOFFFewer trips to the dashboard during a debugging session.
Pricing, straight.
4 TIERS- Verdict on the price
- Fair for testing and for spiky workloads, because the cost estimator and a capped key keep surprises contained.
What people online are saying.
0 QUOTEDToo new for a street reaction: directory listings on Glama and mcpservers.org describe the server, and no public user reviews or discussion threads about it turned up.
The honest part.
4 LIMITATIONSSome capabilities only work once a Zyte API key is attached to your account.
No outside users have published results yet, so reliability on hard anti-bot targets is unproven.
Third-party comparison pages conflict on whether Zyte ships an official MCP server, which muddies vendor research.
Every request the agent makes counts toward your Zyte API limits, including the 3000 RPM standard rate limit.
The field.
3 ALTERNATIVES- Firecrawl MCP
- Official MCP server and CLI for scraping from agents such as Claude Code and Cursor.PICK IT WHENYou want a scraping MCP built mainly around AI coding workflows and plan to compare it head to head.
- Apify MCP
- Exposes Apify Actors to agents through MCP, opening a marketplace of scraping and automation tools.PICK IT WHENYou need prebuilt scrapers for many specific sites beyond what one API covers.
- Bright Data MCP
- Brings search and scraping to agents over MCP with broad anti-bot coverage.PICK IT WHENBlocked targets and access to many data sources matter most.
Getting started.
4 STEPSAdd the server in your client, for example: claude mcp add --transport http zyte https://mcp.zyte.com/v1/mcp
Sign in to Zyte with your client's login command, such as claude mcp login zyte, and sign up there if you are new.
Ask the agent: Fetch https://example.com with Zyte and tell me the title of the page; expect Example Domain and a Zyte MCP tool call in the log.
Set a spending limit on mcp_access_key in the Zyte dashboard before running bigger jobs.
Quick answers.
2 ANSWERS- What happens when the free credit runs out?
- Zyte's pricing page says the account is suspended until you set a spending limit. Setting one before the credit expires keeps the agent working without interruption.
- Can I use it with Cursor or another client that is not in the install list?
- Zyte documents a generic JSON block for most MCP clients that points an http server entry at https://mcp.zyte.com/v1/mcp, after which you sign in through that client. The one-line installs cover Claude Code, Codex CLI, GitHub Copilot CLI, VS Code and claude.ai.
The bottom line.
SPECIMEN 0243 CLOSEDInstall it, run one extraction against a page you know well, and read what that cost before you hand the agent a full crawl.
Field research: 2 pages scraped · 3 search passes · 3 community sources. Reviewed by the Koda desk on 2026-10-07.