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:
-
Name similarity (exact matches always win)
-
Trust score (minimum 7 preferred, 9-10 ideal)
-
Documentation coverage (higher snippet counts)
-
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
-
Think before executing - What does user really need?
-
Resolve first - Unless user provides /org/project
-
Be specific with topics - Multi-word topics work better
-
Scale tokens appropriately - Don't always use maximum
-
Chain related fetches - Build complete context
-
Trust score matters - Prefer libraries with scores ≥7
Quick Reference
-
THINK: What library and documentation needed?
-
RESOLVE: /mcp__context7__resolve-library-id libraryName="[package]"
-
SELECT: Based on trust, snippets, relevance
-
FETCH: /mcp__context7__get-library-docs context7CompatibleLibraryID="[id]" topic="[specific]" tokens="[appropriate]"
-
APPLY: Use documentation to answer question