Documentation Lookup

Retrieve up-to-date documentation links for libraries and frameworks, prioritizing org-level best practices before falling back to external sources.

Workflow

Step 1: Check Org Docs First

Query the "docs" MCP server for library-specific documentation:

1. •Use ListMcpResourcesTool with server: "docs" to list available resources
2. •Look for matching library docs in org://docs/libraries/ (e.g., org://docs/libraries/tanstack-query.md)
3. •If found, use ReadMcpResourceTool to read the org doc

Org docs contain:

• •Best practices and recent updates specific to the organization
• •Links to official documentation pages (often in markdown format)

This info might tell you what docs pages are most relevant to the user's preferences.

Step 2: Parse and Fetch Linked Documentation

When org docs contain documentation links:

1. •If the org docs have a llms.txt link, use that to get the list of all docs pages, and then find the most relevant nested pages/sections for the user task.
2. •Use WebFetch to retrieve each relevant page, to ensure it has the right information for the task.
   • •When fetching URLs from llms.txt, ALWAYS use the .md/.mdx extension exactly as listed. Never strip the extension! Full HTML pages can be 20x larger than the markdown pages.

Step 3: Fallback for Unknown Libraries

** ONLY PERFORM WEB SEARCH AFTER CHECKING ORG DOCS!! **

If the library is NOT found in org docs:

1. •

   First, search for llms.txt: Use WebSearch to find [library-name] llms.txt site:[official-domain] or [library-name] llms.txt

   • •llms.txt files are AI-optimized documentation indexes
   • •If found, fetch and parse for relevant links
2. •

   Otherwise, use Context7 MCP:

   • •Use mcp__context7__resolve-library-id to find the library ID
   • •Use mcp__context7__query-docs with the user's specific question
   • •Extract relevant documentation links from the response

Output Format

Return results with title/description on one line, URL on the next (avoids word wrap issues):

## Documentation for [Library Name]

**Page Title** — Brief description of what this page covers\
https://example.com/docs/page.md

**Page Title** — Section Name\
https://example.com/docs/page.md#section-id

**Getting Started** — Initial setup and configuration
https://library.dev/docs/getting-started.md

**API Reference - useQuery** — Hook API and options
https://library.dev/docs/api.md#useQuery

Each entry should include:

• •Title: The document or page name (bold)
• •Description: Brief context about what the page covers (after em dash)
• •URL: On its own line for easy clicking/copying

Examples

Example 1: User asks about tanstack-query mutations docs

Query: "How do I use mutations in tanstack-query?"

Process:

1. •Check org docs → Find org://docs/libraries/tanstack-query.md
2. •Read org doc → Extract mutation-related links
3. •Fetch linked pages → Find sections about mutations
4. •Return a list of links and page titles

## Documentation for TanStack Query (mutations)

### Sources

**TanStack Query Best Practices** — Org-specific patterns and recommendations
org://docs/libraries/tanstack-query.md

**Mutations** — Creating, updating, and deleting data
https://tanstack.com/query/latest/docs/react/guides/mutations.md

**Mutations - Side Effects** — Running code after mutation success/failure
https://tanstack.com/query/latest/docs/react/guides/mutations.md#mutation-side-effects

**useMutation API Reference** — Hook signature, options, and return values
https://tanstack.com/query/latest/docs/react/reference/useMutation.md

<answer/code-snippets/etc-goes-here>

Example 2: User asks about an unknown library

Query: "How do I set up authentication with Lucia?"

Process:

1. •Check org docs → Not found in org://docs/libraries/
2. •Search (or use a subagent to search) for llms.txt → WebSearch: "lucia auth llms.txt"
3. •If found, parse llms.txt for auth setup links
4. •If not found, use Context7:
   • •Resolve library ID for "lucia"
   • •Query for authentication setup docs
5. •Return relevant links

## Documentation for Lucia (authentication)

### Sources

**Getting Started** — Installation and basic setup
https://lucia-auth.com/getting-started.md

**Sessions** — Creating and managing user sessions
https://lucia-auth.com/sessions.md

**Database Setup** — Configuring adapters for your database
https://lucia-auth.com/database.md

<answer-goes-here>

Key Considerations

Prioritization

1. •ALWAYS check org docs first - they contain organization-specific best practices; DON'T WASTE CONTEXT DOING WEB SEARCHES BEFORE CHECKING ORG DOCS!
2. •When using external sources, prefer llms.txt and corresponding .md/.mdx docs pages over regular .html docs pages; full HTML pages can be 20x larger

Header-Level Links

When the user's query is specific:

• •Fetch the page content
• •Parse headers to find relevant sections
• •Construct fragment URLs (e.g., #getting-started, #useQuery)
• •Include both page-level and section-level links

Handling Multiple Relevant Pages

• •Limit to ~2-5 most relevant links
• •Order by relevance to the user's query

Tools Used

• •ListMcpResourcesTool - List available org docs
• •ReadMcpResourceTool - Read org documentation
• •WebFetch - Fetch external documentation pages
• •WebSearch - Search for llms.txt or documentation
• •mcp__context7__resolve-library-id - Resolve library to Context7 ID
• •mcp__context7__query-docs - Query Context7 for documentation