Mcp Context7

---
context: fork
name: mcp-context7
description: "Provides Context7 MCP server workflow for looking up library documentation, framework patterns, and API references with caching and fallback strategies. Use when user asks about documentation lookup, library usage, framework patterns, import questions, package API, Context7, 라이브러리, or 프레임워크."
lang: [en, ko]
platforms: [claude-code, gemini-cli, codex-cli, cursor]
level: 2
triggers:
  - "documentation"
  - "library"
  - "framework"
  - "import"
  - "package"
  - "Context7"
  - "official docs"
agents:
  - "backend-developer"
  - "frontend-developer"
tokens: "~2K"
category: "tooling"
whenNotToUse: "General reasoning, code review, or business logic tasks that do not require up-to-date library documentation or API reference lookup."
sources:
  - "https://context7.com/docs"
  - "https://modelcontextprotocol.io/introduction"
version: "1.0.0"
lastVerified: "2026-06-08"
source_hash: c6df7ccd
---
# MCP: Context7

## When This Skill Applies
- Implementing features with external libraries
- Looking up framework-specific patterns and conventions
- Verifying API signatures, parameters, options
- Finding official best practices for a technology

## Core Guidance

**Workflow**: resolve-library-id -> query-docs -> apply patterns -> cache results

**Step-by-Step**:
1. Detect library need (import statement, framework question, dependency)
2. Call `resolve-library-id` with library name + user query
3. Call `query-docs` with resolved library ID + specific topic
4. Apply patterns with proper version compatibility
5. Cache successful patterns for session reuse

**Usage Rules**:
- Always resolve library ID first, never guess
- Max 3 calls per question; use best result after 3 attempts
- Be specific: include version and topic in queries
- Verify docs match project dependency versions

**Good Queries**: "How to set up JWT auth in Express.js", "React useEffect cleanup examples", "Next.js App Router dynamic routes TypeScript"

**Bad Queries**: "auth", "hooks", "how to code"

**Error Recovery**:
| Failure | Recovery |
|---------|----------|
| Library not found | Broader search terms -> WebSearch fallback |
| Timeout | Cached knowledge -> note limitations |
| Invalid ID | Retry different terms -> WebSearch |
| Version mismatch | Find compatible version -> suggest upgrade |
| Server down | WebSearch for official docs |

## Quick Reference
- Always: resolve-library-id before query-docs
- Max 3 attempts per question
- Cache results within session
- Fallback: WebSearch -> cached knowledge -> built-in knowledge

## Rationalizations

The following table captures common excuses agents make to skip the rigor of this skill, paired with factual rebuttals.

| Excuse | Rebuttal |
|--------|----------|
| "my training data is recent enough" | library APIs break weekly; Context7 has current docs your training does not |
| "I will just read the README" | READMEs lag behind releases; Context7 queries the live doc source |
| "Context7 is slow" | one Context7 query is faster than debugging a deprecated API call in production |
| "I know this library well" | well enough to know what changed in the last minor version? verify before you commit |
| "caching means stale data" | Context7 caches with TTL and you can force-refresh — stale is a configuration choice |