context7

name: context7 description: "Search documentation using Context7's vector embeddings. Provides semantic search over code documentation (libraries, frameworks, APIs) via the c7 CLI."

Context7 Documentation Search

Use when searching documentation using Context7's vector embeddings. Provides semantic search over code documentation (libraries, frameworks, APIs).

Quick Start

# Search for a project
c7 search react

# Query documentation vectors
c7 mdn fetch API
c7 python requests authentication
c7 nodejs file system

# Get project info
c7 info mdn

# Save results to file
c7 react hooks --save

CLI Reference

Search Projects

Find available documentation projects:

npx context7 search <term>
# or
c7 search <term>

Returns: List of projects with identifiers (paths like mdn/mdn, pandas-dev/pandas, etc.)

Query Vectors

Search documentation using natural language:

npx context7 <projectIdentifier> <query...>
# or  
c7 <project> "how to use hooks"
c7 <project> error handling patterns
c7 <project> async await examples

Parameters:

• projectIdentifier: Project path (mdn/mdn), partial path (react), or unique name
• query: Natural language question or topic

Options:

• -t, --type: Output format (txt | json) — default: txt
• -k, --tokens: Max tokens — default: 5000
• -s, --save: Save output to file — default: false

Project Info

Display metadata about a project:

c7 info <projectIdentifier>

Common Patterns

Quick API Lookup

c7 lodash debounce throttle
c7 axios interceptors error handling
c7 typescript generic constraints

Deep Dive (JSON for parsing)

c7 pandas dataframe filtering -t json -k 10000

Save for Reference

c7 nextjs app router -s -k 8000
# Saves to file in current directory

Multi-Project Search

# Find the right project first
c7 search python data
# Then query the specific one
c7 pandas-dev/pandas groupby aggregation

Project Identifiers

Common patterns:

• mdn/mdn — MDN Web Docs
• facebook/react — React
• microsoft/TypeScript — TypeScript
• python/cpython — Python standard library
• pandas-dev/pandas — Pandas
• numpy/numpy — NumPy

Use c7 search <term> to find the exact identifier.

Tips

• Be specific: "react useEffect cleanup" > "react hooks"
• Include error context: "axios network error handling" > "axios errors"
• Use JSON for piping: -t json | jq '.results[]'
• Token budget: Default 5000 is ~3750 words; increase for deep dives

When to Use

✅ Use Context7 when:

• You need semantic search over docs (not just keyword)
• The topic is API/library usage
• You want curated, high-quality doc sources
• You need code examples with explanations

❌ Don't use when:

• You need real-time/package-specific info (use brave-search)
• The library isn't indexed in Context7
• You need version-specific docs (Context7 may lag)

Usage in pi

Since pi doesn't have MCP client support, use the bash tool:

# Basic query
bash: c7 mdn fetch API

# Save JSON for processing
bash: c7 pandas groupby -t json -s
read: pandas-groupby.json

# Search then query (two-step)
bash: c7 search numpy
bash: c7 numpy array broadcasting -k 3000

Error Handling

Parsing JSON Output

Save JSON, then extract code examples:

# Query and save JSON
bash: c7 typescript generics -t json -s -k 2000

# Extract all code blocks
bash: uv run ${CLAUDE_SKILL_DIR}/scripts/extract_code.py typescript-generics.json

# Filter by language
bash: uv run ${CLAUDE_SKILL_DIR}/scripts/extract_code.py typescript-generics.json -l typescript

# Output as JSON for further processing
bash: uv run ${CLAUDE_SKILL_DIR}/scripts/extract_code.py typescript-generics.json -f json

# Pipe directly (no file)
bash: c7 pandas groupby -t json | uv run ${CLAUDE_SKILL_DIR}/scripts/extract_code.py -l python

extract_code.py options: