context7（Context7 文档搜索）

Context7 Documentation Retrieval

Framework for using /mcp__context7__resolve-library-id and /mcp__context7__get-library-docs tools effectively.

Process: THINK → RESOLVE → FETCH → APPLY

Command: /mcp__context7__resolve-library-id

Resolves a package name to a Context7-compatible library ID.

Parameters

• libraryName (required): Library name to search for

Returns

List of matching libraries (10-30+ results) with:

• Context7-compatible library ID: Format /org/project
• Trust Score: Authority indicator (1-10, higher is better)
• Code Snippets: Number of available examples
• Description: Library summary

Selection Process

Priority order:

1. Name similarity (exact matches always win)
2. Trust score (minimum 7 preferred, 9-10 ideal)
3. Documentation coverage (higher snippet counts)
4. Description relevance to query intent

Decision rules:

• Exact name + trust ≥7 → Select it
• Multiple high-trust (≥9) → Choose highest snippet count
• Low trust (<7) → Prioritize snippet count
• Ambiguous → Request user clarification

Skip resolve when:

• User provides path like /vercel/next.js
• Query contains /org/project format

Always resolve when:

• User mentions name only: "React", "Next.js"
• Generic terms: "React docs"

Common Patterns

"react" → /reactjs/react.dev (trust: 10)
"next.js" → /vercel/next.js (trust: 9.5)
"vue" → /vuejs/core (trust: 10)
"mongodb" → /mongodb/docs (trust: 9.8)

Command: /mcp__context7__get-library-docs

Fetches documentation using exact Context7-compatible library ID.

Parameters

• context7CompatibleLibraryID (required): Exact ID from resolve step
• topic (optional): Specific focus area
• tokens (optional): Max tokens (default: 10000, range: 1000-20000)

Token Strategy

• 3K-5K: Quick reference (single function/method)
• 6K-10K: Feature exploration
• 10K-15K: Implementation guides
• 15K-20K: Comprehensive learning

Topic Selection

Be specific with multi-word topics:

• ❌ "api" → ✅ "REST API endpoints"
• ❌ "hooks" → ✅ "useState useEffect lifecycle"
• ❌ "css" → ✅ "responsive design breakpoints"

Workflows

React Implementation

THINK: Need React infinite scroll docs
RESOLVE: /mcp__context7__resolve-library-id libraryName="react"
SELECT: /reactjs/react.dev (trust: 10)
FETCH: /mcp__context7__get-library-docs context7CompatibleLibraryID="/reactjs/react.dev" topic="infinite scroll virtualization" tokens=12000

Next.js Debugging

THINK: Debug hydration errors
RESOLVE: /mcp__context7__resolve-library-id libraryName="next.js"
SELECT: /vercel/next.js (trust: 9.5)
FETCH: /mcp__context7__get-library-docs context7CompatibleLibraryID="/vercel/next.js" topic="hydration errors debugging SSR" tokens=15000

Direct ID Usage

THINK: User provided /mongodb/docs
FETCH: /mcp__context7__get-library-docs context7CompatibleLibraryID="/mongodb/docs" topic="aggregation pipeline" tokens=15000

Error Handling

No matches: Try alternative names (vue.js → vue) Multiple matches: Ask user preference Wrong library: Re-run with different term Insufficient docs: Increase tokens or refine topic

Best Practices

1. Think before executing - What does user really need?
2. Resolve first - Unless user provides /org/project
3. Be specific with topics - Multi-word topics work better
4. Scale tokens appropriately - Don't always use maximum
5. Chain related fetches - Build complete context
6. Trust score matters - Prefer libraries with scores ≥7

Quick Reference

1. THINK: What library and documentation needed?
2. RESOLVE: /mcp__context7__resolve-library-id libraryName="[package]"
3. SELECT: Based on trust, snippets, relevance
4. FETCH: /mcp__context7__get-library-docs context7CompatibleLibraryID="[id]" topic="[specific]" tokens="[appropriate]"
5. APPLY: Use documentation to answer question