What is the Context7 MCP server, and what does it not do for a team?

Short answer

Context7, from Upstash, is an MCP server that gives coding agents current, version-specific documentation and code examples for public libraries, so they rely less on outdated training data. It has two tools, resolve-library-id and query-docs, and runs remotely at https://mcp.context7.com/mcp. As of November 2026 the Free plan includes 1,000 API calls a month and Pro costs $10 per seat a month, with private repositories.

What Context7 does

A coding agent’s knowledge of a library stops at its training date. Ask it about a framework that changed its API last quarter and it may write code for the old version. Context7 fills that gap: it indexes documentation and examples for public libraries, and lets the agent fetch the right version at the moment it needs it.

It does that with two MCP tools:

  1. resolve-library-id turns a name like “supabase” into a Context7 library ID.
  2. query-docs fetches documentation for that ID and your question, optionally for a specific language.

The agent calls the first, then the second, and the result lands in its context. You can skip the first step by naming the library ID yourself.

What Context7 indexes is documentation, not code. It parses .md, .rst and .html files, converts PDF and DOCX to text, skips raw source files when documentation is present, and skips pages whose prose is mainly in another language. A library owner can set included folders and exclusions with a context7.json file, and Context7 refreshes documentation automatically based on popularity, or on every push to the default branch with its GitHub Action.

Setting it up

Option How
Remote MCP server URL https://mcp.context7.com/mcp, API key in an Authorization: Bearer header
Local MCP server npm package @upstash/context7-mcp
CLI with skills, no MCP npx ctx7 setup, then ctx7 library and ctx7 docs (Node.js 22.13+)

A free API key from the Context7 dashboard raises the rate limits. For a team, add the remote server once in the shared project config: .mcp.json for Claude Code, .cursor/mcp.json for Cursor, config.toml for Codex. See MCP servers in Claude Code for a team and Cursor MCP servers.

Plans, as of November 2026

Plan Price API calls Private repos
Free $0 1,000 a month, then 20 bonus calls a day until the month resets No
Pro $10 per seat a month 2,000 per seat a month, then $5 per 1,000 Yes, parsing $5 per 1M tokens
Enterprise Custom Custom Yes

On Pro, each member has their own allocation; calls are not pooled.

The common misunderstanding

“With Context7 the agent knows our stack.” It knows the public documentation of the libraries in your stack. That is different from knowing how your team uses them. Context7 has no way of knowing that you wrap the database client in your own module, that you never use a certain ORM feature since an incident, or that your reviewers reject a pattern the official docs recommend. Those are your team’s rules, and they are usually not in any public library.

Private repository parsing on Pro gets closer, because the agent can query your own code as documentation. But code shows what you did, not why, and not what a reviewer corrected last week.

Also keep in mind what Context7 says about itself: the documentation is community-contributed, and Context7 cannot guarantee its accuracy, completeness or security. Anything the agent reads lands in its context, so it is worth treating it like any other external input. More on that in is MCP safe to use in a company.

What this means for a team

Context7 solves a real, narrow problem: stale knowledge of public libraries. It is a good default for teams that build on fast-moving frameworks. It does not solve the other half of what makes a coding agent useful in your codebase: the conventions, decisions and corrections that are specific to your team. That half has to come from somewhere else, and today it usually sits in CLAUDE.md, AGENTS.md or rules files that each tool reads differently. See shared memory across coding agents.

Where Bombos fits

Context7 brings public library docs to the agent; Bombos brings your team’s own knowledge. It is a plugin that gives Claude Code, Cursor, Codex, Copilot and your other AI tools one shared memory with team knowledge, rules the AI learned from corrections, and what already sits in your repos, documents and mail, with a trail of what each answer rests on. The two sit side by side. It runs in our cloud or on your own server, and one developer can try it first with their own agents.

Related questions

How do I make the AI use Context7?

Add "use context7" to your prompt, or name a library directly, for example "use library /supabase/supabase". You can also put a rule in CLAUDE.md or your Cursor rules so the agent uses it without being asked.

Is Context7 free?

There is a Free plan with 1,000 API calls a month. After that, calls are blocked except for 20 bonus calls a day until the month resets. Pro is $10 per seat a month with 2,000 calls per seat included.

Can Context7 read our private repositories?

On Pro and Enterprise, yes. Parsing private repositories costs $5 per million tokens, according to Context7's plans page.

Is the documentation in Context7 checked?

Context7 says the documentation is community-contributed and that it cannot guarantee its accuracy, completeness or security. Projects are maintained by their owners, and you can report suspicious content.

Do I need MCP to use Context7?

No. There is also a CLI with skills: ctx7 library and ctx7 docs, set up with npx ctx7 setup. It needs Node.js 22.13 or newer.

Sources

  1. upstash/context7 (GitHub README)
  2. Context7 plans
  3. Adding Libraries (Context7 docs)
The first half hour · free

What does your AI forget?

  1. 01You tell us which AI you use and where it keeps starting over.
  2. 02We show what a shared memory changes there, on your own examples.
  3. 03You keep the plan, with or without Bombos.
Piet Baudoin
Piet BaudoinYou talk to Piet, from the Bombos teamDutch team from Utrecht and Nijmegen · data in Europe
We reply the same working day.