Upstash is the serverless data platform for modern and AI applications. Its team also builds Context7, which gets current, version-specific documentation and code examples into Claude, Codex, Cursor, and other coding agents.
Behind Context7 is a constant crawl of documentation sites. Context.dev powers that crawling, handling millions of requests every day so the docs Context7 serves stay current.
Fresh docs are the product
Coding agents are only as good as the docs they read. Training data falls behind every release, so an agent asked about a new version of a library can confidently suggest methods that no longer exist. Context7 closes that gap by indexing library documentation and handing the agent the right version at the moment it writes code.
That only works if the index keeps up. Every library Context7 covers has a docs site that needs crawling, and those sites change as new versions ship. At Context7's volume, the crawling layer has to be fast, reliable, and consistent across very different documentation sites.
Set up by Claude in minutes
Abdullah Enes Güleş, a software engineer at Upstash building Context7, didn't wire up the integration by hand. Claude did it for him:
"Incredible AX! Claude handled all the setup for me."
AX is agent experience: how easily an AI agent can discover, understand, and integrate an API. For a team whose product exists to help agents read docs, an API that agents can set up on their own was a natural fit.
Benchmarked against the previous crawler
With the integration in place, the team went straight to a head-to-head comparison with its previous crawling service:
"It wasn't longer than a few minutes until we started a few benchmarks with our previous crawling service and already saw the benefits and improvements."
The improvements showed up in the first benchmarks, and Context.dev now runs the website crawling Context7 uses to index docs.
Millions of requests a day
Context7 crawls documentation on Context.dev at a scale of millions of requests per day. That volume is what keeps the index current, so the docs an agent pulls into Claude, Codex, or Cursor reflect what a library actually ships today.
Building docs search or a knowledge base for AI agents? Explore web crawling with Context.dev, RAG pipelines, and the Web Scraping API.