Auto-discover and fetch public CLAUDE.md files from products/libraries

Preflight Checklist

• [x] I have searched existing requests and this feature hasn't been requested yet
• [x] This is a single feature request (not multiple features)

Problem Statement

Claude Code only reads CLAUDE.md files that exist in the local codebase. When users ask for help with external products or libraries, Claude doesn't automatically check if those products have published agent instructions.

Proposed Solution

When a user asks for help with a product/library, Claude Code should automatically attempt to discover and fetch public CLAUDE.md files from known locations:

1. GitHub repositories: https://github.com/{org}/{repo}/CLAUDE.md or https://github.com/{org}/{repo}/blob/main/CLAUDE.md
2. Documentation sites: https://docs.{product}.io/CLAUDE.md
3. Product websites: https://{product}.com/CLAUDE.md
4. MCP registry or similar: A registry of known CLAUDE.md locations

Alternative Solutions

_No response_

Priority

Medium - Would be very helpful

Feature Category

Configuration and settings

Use Case Example

User: "Help me use Pinecone for semantic search"

Current behavior: Claude provides general help based on training data

Proposed behavior:

1. Claude detects "Pinecone" is an external product
2. Attempts to fetch https://docs.pinecone.io/CLAUDE.md (or other known locations)
3. If found, uses those instructions to provide accurate, up-to-date guidance
4. Falls back to general knowledge if not found

Benefits

1. Automatic accuracy: Users get product-specific best practices without manual intervention
2. Up-to-date guidance: Product teams can update their CLAUDE.md files independently of Claude's training
3. Reduced friction: No need for users to manually point Claude to documentation
4. Ecosystem growth: Encourages products to publish AI assistant instructions
5. Standardization: Creates a convention for machine-readable product documentation

Technical Considerations

• Caching: Cache fetched files for the session to avoid repeated requests
• Fallback: Gracefully handle missing files without disrupting workflow
• Security: Only fetch from trusted domains or verified sources
• Privacy: Make discovery opt-in/configurable if needed
• Performance: Parallel/async fetching to avoid blocking
• Versioning: Consider supporting versioned CLAUDE.md files (e.g., CLAUDE-v2.md)

Real-World Example

In the Pinecone documentation repository, there's already an assets/AGENTS-PYTHON.md file with detailed instructions for AI agents. If this were published as a public CLAUDE.md file, Claude could automatically use it when helping users with Pinecone, ensuring:

• Correct method names (e.g., upsert_records() not upsert())
• Current API patterns
• Product-specific best practices
• Up-to-date quickstart flows

Additional Context

_No response_