Context7 launches Docs7: documentation served to agents, not just to human readers
Upstash's Context7 has launched Docs7, a documentation platform that, the company says, serves one source as both a human site and agent-readable formats, with analytics on which agents read it.
Context7, a project by the Upstash team, announced Docs7 on 3 September: a documentation platform aimed first at the coding agents that read documentation on people's behalf, while the same source still publishes the site humans read. The post, from an X account with 8,329 followers, calls it "the best way to serve documentation to agents" and promises clean, agent-readable docs, native WebMCP and skills support, and insight into AI traffic and searches. It drew 30,277 views.
How AI-ready is your URL?
The announcement video shows a four-item checklist. A URL passes if it:
- Serves Markdown for Accept: text/markdown (when a client asks for it)
- Sets Vary: Accept (the header saying the answer varies by request)
- Rejects unsupported types with 406 (the refusal code for an unsupported format)
- Honors q-values (the weights a client puts on its preferences)
These are ordinary HTTP content negotiation rules. The company's framing: agents want clean structure and runnable examples, not rendered HTML. The product page says one source produces both the site people read and the formats agents use, so nobody keeps a second copy: HTML pages, Markdown pages, an llms.txt file and a search API.
Documentation as a measured channel
The other half is measurement. The product page says Agent Analytics shows which agent read the docs, what it searched for and what it never found, and that the Autonomous Docs Agent turns every unanswered Context7 query into a pull request. The company says an existing Mintlify repository works as it stands, same docs.json, same MDX, same components, and describes migration as a separate step: connect the repository, take a preview, then point the domain at Docs7, with canonical data, sitemap.xml, robots.txt and permanent redirects generated for SEO.
Free costs 0 dollars: custom domain, Agent Analytics, the Context7 integration, WebMCP, the Search API and a Playground. Pro is 200 dollars a month, adding autonomous docs agents, a preview deployment per pull request, an assistant, team support and SOC2 certification. The product page reports more than 100 million agent requests all time, over 18 million in the last 30 days and 122,000 indexed libraries; these are the company's own descriptions and figures, none of it independently verified. See the announcement and the product page.
