# Context7 MCP Server: Up-to-Date Docs for Your AI Coding Assistant (Setup Guide)

> Context7 is an MCP server that fetches up-to-date library documentation so your AI assistant can reference current APIs instead of relying on training data. Setup guide for Cursor, Claude Code, VS Code, and more.

**Source:** https://mcpplaygroundonline.com/blog/context7-mcp-server-setup-guide  
**Author:** Nikhil Tiwari  
**Published:** 2026-02-08  
**Updated:** 2026-02-08  
**Category:** Tutorial  
**Reading time:** 10 min read

---

TL;DR

-   **Context7** is an MCP server that fetches up-to-date, version-specific library documentation for AI assistants
-   Helps reduce outdated code and hallucinated APIs by adding current docs to the model's context
-   Add `use context7` to your prompt to trigger a documentation lookup
-   Works with Cursor, Claude Code, VS Code, Windsurf, and other MCP clients
-   Open source by Upstash — MIT license ([GitHub](https://github.com/upstash/context7))

A common frustration with AI coding assistants: you ask for code using a library, and the response is based on an outdated API or a function that no longer exists. The model's training data has a cut-off, and libraries move fast.

[Context7](https://github.com/upstash/context7) is an MCP server designed to help with this. It fetches up-to-date, version-specific documentation and code examples from indexed library sources and adds them to your AI assistant's context window. The idea is straightforward: give the model current docs so it has a better chance of generating correct code.

## The Problem Context7 Solves

**Without Context7**

-   Code may rely on outdated training data
-   Risk of hallucinated APIs that don't exist
-   Answers may target the wrong package version
-   Manual verification often needed

**With Context7**

-   Docs fetched from indexed library sources
-   Version-specific examples when available
-   Model has current API surface in context
-   Triggered by adding "use context7" to your prompt

## Quick Facts

Detail

Info

**GitHub**

[upstash/context7](https://github.com/upstash/context7)

**npm**

`@upstash/context7-mcp`

**Remote URL**

`https://mcp.context7.com/mcp`

**License**

MIT

**Built by**

[Upstash](https://upstash.com)

**API key**

Free at [context7.com/dashboard](https://context7.com/dashboard) (higher rate limits)

**Docs**

[context7.com/docs](https://context7.com/docs)

## How It Works

Context7 exposes two MCP tools. When you include "use context7" in a prompt, the AI assistant calls them as part of its tool-use flow:

**1\. resolve-library-id**

Searches for a library by name (e.g., "nextjs", "react") and returns a Context7 library ID. Ranked by name match, description relevance, snippet count, and trust score.

**2\. get-library-docs**

Fetches the actual documentation and code examples for a resolved library ID. Returns version-specific content from the source.

Typical flow: your prompt mentions a library → the AI calls `resolve-library-id` → gets the library ID → calls `get-library-docs` → the returned documentation is added to context → the model uses it when generating a response.

## Setup: Cursor

Edit your Cursor MCP config file at `~/.cursor/mcp.json`:

### Option A: Remote HTTP server (simpler, no local process)

```
{
 "mcpServers": {
 "context7": {
 "url": "https://mcp.context7.com/mcp",
 "headers": {
 "CONTEXT7_API_KEY": "YOUR_API_KEY"
 }
 }
 }
}
```

### Option B: Local stdio server via npx

```
{
 "mcpServers": {
 "context7": {
 "command": "npx",
 "args": ["-y", "@upstash/context7-mcp"],
 "env": {
 "CONTEXT7_API_KEY": "YOUR_API_KEY"
 }
 }
 }
}
```

An API key is optional but recommended for higher rate limits. Get one free at [context7.com/dashboard](https://context7.com/dashboard). After saving the config, restart Cursor for the server to load.

## Setup: Claude Code

### Option A: Remote HTTP server

```
claude mcp add --transport http \
 --header "CONTEXT7_API_KEY: YOUR_API_KEY" \
 context7 https://mcp.context7.com/mcp
```

### Option B: Local stdio server via npx

```
claude mcp add --transport stdio \
 --env CONTEXT7_API_KEY=YOUR_API_KEY \
 context7 -- npx -y @upstash/context7-mcp
```

Verify the server is registered:

```
claude mcp list
```

You can also check status inside Claude Code with `/mcp`.

## Setup: VS Code (Copilot Chat)

Add to `.vscode/mcp.json` in your project or your VS Code user settings:

```
{
 "servers": {
 "context7": {
 "type": "http",
 "url": "https://mcp.context7.com/mcp",
 "headers": {
 "CONTEXT7_API_KEY": "YOUR_API_KEY"
 }
 }
 }
}
```

## Setup: Claude Desktop

Claude Desktop supports stdio servers only. Add to your `claude_desktop_config.json`:

```
{
 "mcpServers": {
 "context7": {
 "command": "npx",
 "args": ["-y", "@upstash/context7-mcp"],
 "env": {
 "CONTEXT7_API_KEY": "YOUR_API_KEY"
 }
 }
 }
}
```

Restart Claude Desktop after saving. The hammer icon in the text input indicates tools are loaded.

## Usage

Once installed, just add **"use context7"** to any prompt:

```
# Examples
"Create a Next.js middleware that checks for a valid JWT. use context7"

"Set up a Hono API route with Zod validation. use context7"

"Configure a Cloudflare Worker to cache API responses for 5 minutes. use context7"

"Write a React Server Component that fetches data with Suspense. use context7"
```

What happens behind the scenes:

1.  The AI identifies the library mentioned in your prompt
2.  It calls `resolve-library-id` to find the library in Context7's index
3.  It calls `get-library-docs` to fetch relevant documentation
4.  The returned docs are added to context, and the model uses them when generating its response

## Advanced Features

### Private Repositories

Paid plans support private repositories. You can add them through the [Context7 dashboard](https://context7.com/dashboard) so the server can index and serve documentation for internal libraries.

### Library Verification

Library owners can [claim and verify their libraries](https://context7.com/docs/howto/claiming-libraries) on Context7. Verified libraries are flagged with a higher trust score in search results.

### Direct API Usage

Context7 also has a REST API for programmatic access:

```
# Search for a library
curl -H "Authorization: Bearer YOUR_API_KEY" \
 "https://context7.com/api/v2/libs/search?query=nextjs"

# Get documentation
curl -H "Authorization: Bearer YOUR_API_KEY" \
 "https://context7.com/api/v2/context?libraryId=/vercel/next.js"
```

### TypeScript SDK

For building custom integrations, Context7 provides a TypeScript SDK and a CLI tool. See the [API Guide](https://context7.com/docs/api-guide) for details.

## Why Use Context7 Over Just Searching Docs?

Approach

Pros

Cons

**LLM training data**

No setup needed

Has a knowledge cut-off; may hallucinate or use old APIs

**Manual doc search**

Authoritative, current

Slow, requires context-switching

**Custom RAG pipeline**

Fully customizable, works with any source

Significant setup and maintenance effort

**Context7 MCP**

Quick setup, docs fetched on demand

Library must be in Context7's index; quality varies by library

## Supported Clients

Context7 should work with any MCP-compatible client. The following have documented setup guides on [context7.com](https://context7.com/docs/resources/all-clients):

**Cursor**

**Claude Code**

**Claude Desktop**

**VS Code**

**Windsurf**

**OpenCode**

Test MCP Servers in Your Browser

Try Context7 and other remote MCP servers with MCP Playground

[Open MCP Playground →](/mcp-test-server)

## Related Content

-   [What Is the Model Context Protocol (MCP)?](/blog/what-is-model-context-protocol)
-   [Claude Code MCP: Best Servers Setup Guide](/blog/claude-code-mcp-setup-best-servers-guide)
-   [Awesome MCP Servers List](/blog/awesome-mcp-servers)
-   [Build Your First MCP Server with Python and FastMCP](/blog/build-mcp-server-python-fastmcp-tutorial)

## Frequently Asked Questions

**Is Context7 free?**

There is a free tier that works without an API key, subject to rate limits. A free API key (available at [context7.com/dashboard](https://context7.com/dashboard)) raises those limits. Paid plans add features like private repository indexing and team management.

**What libraries does Context7 support?**

Context7 indexes a large number of open-source libraries, but coverage is not exhaustive. Popular libraries (React, Next.js, Python packages, etc.) are generally well-covered. If a library you need is missing, library owners can [claim and add](https://context7.com/docs/howto/claiming-libraries) their libraries. Private repositories can be added on paid plans.

**Should I use the remote server or local npx?**

The remote HTTP server (`https://mcp.context7.com/mcp`) is simpler — no local process to manage. The local npx option (`@upstash/context7-mcp`) runs on your machine, which may be preferable behind corporate firewalls or for lower latency. Both connect to the same Context7 backend for documentation data.

**Does "use context7" work automatically or do I have to type it every time?**

You need to include "use context7" in your prompt to trigger the documentation lookup. Without it, the AI uses its training data as usual. Some users add it to their system prompt or Cursor rules so it's always active.

**Can I use Context7 with ChatGPT?**

If you're using the OpenAI Agents SDK, you can add Context7's remote URL as a Streamable HTTP or hosted MCP tool. For ChatGPT Apps (available on Business/Enterprise plans), you could configure the endpoint as an MCP server — though we haven't verified this specific combination. See our [ChatGPT MCP setup guide](/blog/chatgpt-mcp-openai-agents-sdk-setup-guide) for general instructions.

---

_Canonical page: https://mcpplaygroundonline.com/blog/context7-mcp-server-setup-guide — MCP Playground (mcpplaygroundonline.com), the free browser-based tool for testing MCP servers and building AI agents._