The archive · Developer & Business Tools · Product decision · 2021–2026
Upstash bets AI agents need fresh code docs; Context7 hits 20.9K GitHub stars in a quarter
Upstash's free Context7 MCP server injects current, version-specific library docs into AI coding prompts — the fastest-growing MCP repo of Q3 2025.
Upstash
What the business is
Serverless Redis, Kafka, QStash and Vector data services priced by consumption, plus Context7: an open-source MCP server and CLI that feeds current, version-specific library docs into AI prompts.
Starting capital：$1.9M seed (2022); $10M Series A led by Andreessen Horowitz (Feb 2024) — about $11.9M total.
How it started
Upstash was founded in 2021 by Enes Akar to sell serverless Redis and Kafka with pay-when-you-use pricing, and took a $10M Series A from a16z in February 2024 as AI apps flooded its infrastructure. Context7, launched in 2025, attacked the pain those developers hit daily: LLMs generating code against outdated or hallucinated APIs.
What happened
Context7 grew from about 1K to 20.9K GitHub stars in Q3 2025 (20.9x, Runa Capital ROSS Index), becoming one of the most-installed MCP servers for coding agents. Upstash expanded it into per-agent plugins for Claude Code, Cursor, Codex and Copilot, and in August 2026 shipped a version in the new Agent Plugins format alongside Vercel, AWS, Cursor, GitHub, Microsoft and OpenAI.
How it ended up
Live and scaling: as of September 2026 Context7 remains free and MIT-licensed with the MCP package at v4.x, serving docs for thousands of libraries into the major coding agents, while its API backend, parsing engine and crawling engine stay proprietary inside Upstash's paid platform.
Background
Context7 is Upstash's open-source answer to one of AI coding's most visible failures: LLMs generating code against stale or invented library APIs. It works as an MCP server or CLI that pulls current, version-specific documentation and code examples from the source and drops them into the prompt, so a coding agent answers with the API that actually exists.
Upstash itself is a serverless data platform — Redis, Kafka, QStash and a vector database — founded in 2021 by Enes Akar with consumption-based pricing, a $1.9M seed in 2022 and a $10M Series A led by Andreessen Horowitz in February 2024. Context7 was the company's bet that AI agents' documentation problem would become the front door to its platform.
The bet paid off in distribution: Context7 grew from about 1K to 20.9K GitHub stars by 20 July 2025 (20.9x in Q3 2025, per Runa Capital's ROSS Index), and by August 2026 Upstash shipped separate plugins for Claude Code, Cursor, Codex and Copilot, plus a version in the new Agent Plugins format. The MCP package sits at v4.x.
Context7 stays free and MIT-licensed while its API backend, parsing engine and crawling engine remain proprietary — an open-source wedge in front of a paid platform, with the GitHub star count as both marketing and moat.
What has to be true
- Docs are the highest-frequency failure point for coding agents, so fixing them creates daily value and word of mouth among exactly the developers Upstash sells to.
- Open-sourcing the MCP server turned every install into distribution: one repo, 20.9K stars in a quarter, no ad spend.
- The MCP standard made one docs server portable across Cursor, Claude Code, Codex and Copilot, multiplying reach without rebuilding the product per editor.
- Keeping the crawler and parsing engine proprietary preserved the paid moat behind the free wedge.
What can be applied
A free tool that fixes the hottest platform's most common failure can beat advertising: Context7's star growth put Upstash in front of exactly the developers it sells to.
Aftermath
As of September 2026 Context7 is live and expanding: the MCP package is at v4.x, plugins cover Claude Code, Cursor, Codex and Copilot, and in August 2026 Upstash shipped a version in the Agent Plugins format. The GitHub repo remains MIT-licensed and one of the most-starred MCP servers, while the private API, parsing and crawling engines anchor Upstash's paid platform — the company continues to sell serverless Redis, Kafka, QStash and vector storage on top of the developer attention Context7 brings in.
Sources
- Agent Plugins Explained: How We Built the Context7 Plugin
- Upstash's serverless data platform hits ARR of $1M just two years after seed funding
- Top trending open-source startups in Q3 2025
- upstash/context7
spotted an error? The archive wants to know.
Your turn
You just read one. Describe what you are building, and see who is betting on the same thing.
Free account · 3 free questions · no card