feat: integrate llms.txt documentation tools

modelcontextprotocol/README.md

@@ -137,6 +137,8 @@ Open `Cursor > Settings > Tools & Integrations > New MCP Server` to include the
 
 ## Available Tools
 
+### Atlan Asset Management Tools
+
 | Tool                | Description                                                       |
 | ------------------- | ----------------------------------------------------------------- |
 | `search_assets`     | Search for assets based on conditions                             |
@@ -147,6 +149,27 @@ Open `Cursor > Settings > Tools & Integrations > New MCP Server` to include the
 | `create_glossary_categories` | Create glossary categories                               |
 | `create_glossary_terms` | Create glossary terms                                         |
 
+### Documentation Tools
+
+| Tool                | Description                                                       |
+| ------------------- | ----------------------------------------------------------------- |
+| `documentation_tool`| Unified tool for Atlan documentation operations with actions: `list_sources`, `fetch_content` |
+
+> **📖 Documentation Tool Usage**: This unified tool enables AI agents to access and retrieve Atlan product documentation to answer user questions about features, APIs, integrations, and usage instructions. It automatically handles domain security, content parsing with optimized HTML-to-markdown conversion, and provides a streamlined interface for all documentation operations. The tool achieves up to 88% token reduction while preserving all important content and links.
+
+#### Documentation Tool Actions
+
+| Action | Description | Parameters | Use Cases |
+|--------|-------------|------------|-----------|
+| `list_sources` | Lists all available Atlan documentation sources with their llms.txt URLs and allowed domains | None | Discover what documentation sources are available |
+| `fetch_content` | Fetches content from any valid Atlan documentation URL | `url` (required), `source_names` (optional) | Get llms.txt index, retrieve specific documentation pages, access nested documentation content |
+
+#### Common Documentation Workflows
+
+1. **Discovery**: Use `list_sources` to see available documentation sources
+2. **Index Retrieval**: Use `fetch_content` with an llms.txt URL to see all available topics
+3. **Content Fetching**: Use `fetch_content` with specific documentation URLs to get detailed content
+
 ## Tool Access Control
 
 The Atlan MCP Server includes a configurable tool restriction middleware that allows you to control which tools are available to users. This is useful for implementing role-based access control or restricting certain operations in specific environments.
@@ -204,6 +227,7 @@ You can restrict access to specific tools using the `RESTRICTED_TOOLS` environme
 
 You can restrict any of the following tools:
 
+#### Asset Management Tools
 - `search_assets_tool` - Asset search functionality
 - `get_assets_by_dsl_tool` - DSL query execution
 - `traverse_lineage_tool` - Lineage traversal
@@ -212,6 +236,9 @@ You can restrict any of the following tools:
 - `create_glossary_categories` - Category creation
 - `create_glossary_terms` - Term creation
 
+#### Documentation Tools
+- `documentation_tool` - Unified Atlan documentation tool (actions: list_sources, fetch_content)
+
 ### Common Use Cases
 
 #### Read-Only Access

modelcontextprotocol/pyproject.toml

@@ -18,6 +18,8 @@ classifiers = [
 dependencies = [
     "fastmcp==2.11.0",
     "pyatlan>=6.0.1",
+    "httpx>=0.25.0",
+    "markdownify>=0.11.6",
 ]
 
 [project.scripts]
@@ -39,3 +41,8 @@ packages = ["."]
 [build-system]
 requires = ["hatchling"]
 build-backend = "hatchling.build"
+
+[dependency-groups]
+dev = [
+    "pre-commit>=4.3.0",
+]

modelcontextprotocol/server.py

@@ -11,6 +11,8 @@
     create_glossary_category_assets,
     create_glossary_assets,
     create_glossary_term_assets,
+    list_doc_sources,
+    fetch_documentation,
     UpdatableAttribute,
     CertificateStatus,
     UpdatableAsset,
@@ -694,6 +696,103 @@ def create_glossary_categories(categories) -> List[Dict[str, Any]]:
     return create_glossary_category_assets(categories)
 
 
+@mcp.tool()
+async def documentation_tool(
+    action: str, url: str = None, source_names: str = None
+) -> Dict[str, Any]:
+    """
+    UNIFIED DOCUMENTATION TOOL - Your one-stop solution for all Atlan documentation needs!
+
+    WHEN TO USE: Call this tool whenever users ask questions about Atlan products, features,
+    integrations, APIs, usage instructions, or any other Atlan-related topics.
+
+    This powerful unified tool handles all Atlan documentation operations through different actions:
+    - list_sources: Discover available Atlan documentation sources with their llms.txt URLs
+    - fetch_content: Retrieve Atlan documentation content from any valid URL
+
+    Args:
+        action (str): The operation to perform. Must be one of:
+            - "list_sources": List all available Atlan documentation sources with llms.txt URLs
+            - "fetch_content": Fetch content from any Atlan documentation URL
+
+        url (str, optional): Documentation URL to fetch content from (for fetch_content)
+        source_names (str, optional): Comma-separated source names for domain checking (for fetch_content)
+
+    Returns:
+        Dict[str, Any]: Response varies by action:
+
+        list_sources: List of available Atlan sources with names, llms.txt URLs, allowed domains, and descriptions
+        fetch_content: Documentation content, success status, and metadata
+
+    FETCH_CONTENT USE CASES:
+    The fetch_content action is versatile and supports multiple scenarios:
+
+    1. **Fetch llms.txt index**: Get all available documentation URLs
+       - Use the llms.txt URL from list_sources to see what topics are available
+       - This returns the raw llms.txt content with all documentation URLs
+
+    2. **Fetch specific documentation pages**: Read any documentation webpage
+       - Main documentation URLs listed in llms.txt
+       - Nested/linked pages within the documentation site
+       - Any docs.atlan.com URL for detailed content
+
+    Examples:
+        # 1. Discover available Atlan documentation sources and their llms.txt URLs
+        await documentation_tool(action="list_sources")
+
+        # 2. Fetch llms.txt to see all available documentation topics
+        await documentation_tool(
+            action="fetch_content",
+            url="https://docs.atlan.com/llms.txt",
+            source_names="Atlan Docs"
+        )
+
+        # 3. Fetch specific documentation page to answer user questions
+        await documentation_tool(
+            action="fetch_content",
+            url="https://docs.atlan.com/integrations/snowflake",
+            source_names="Atlan Docs"
+        )
+
+        # 4. Fetch any nested documentation page
+        await documentation_tool(
+            action="fetch_content",
+            url="https://docs.atlan.com/setup-and-settings/admin/lineage-settings",
+            source_names="Atlan Docs"
+        )
+
+    Security Features:
+    - Domain validation prevents access to unauthorized websites
+    - Source management ensures only trusted Atlan documentation sources
+    - Comprehensive error handling for network and parsing issues
+    """
+
+    if action == "list_sources":
+        return {"sources": list_doc_sources(), "action": "list_sources"}
+
+    elif action == "fetch_content":
+        if not url:
+            return {
+                "error": "url is required for fetch_content action",
+                "action": "fetch_content",
+            }
+
+        # Parse source_names if provided
+        parsed_source_names = None
+        if source_names:
+            parsed_source_names = [name.strip() for name in source_names.split(",")]
+
+        result = await fetch_documentation(url, parsed_source_names)
+        result["action"] = "fetch_content"
+        return result
+
+    else:
+        return {
+            "error": f"Invalid action '{action}'. Must be one of: list_sources, fetch_content",
+            "available_actions": ["list_sources", "fetch_content"],
+        }
+
+
 def main():
     mcp.run()
 

modelcontextprotocol/tools/__init__.py

@@ -7,6 +7,10 @@
     create_glossary_assets,
     create_glossary_term_assets,
 )
+from .docs import (
+    list_doc_sources,
+    fetch_documentation,
+)
 from .models import (
     CertificateStatus,
     UpdatableAttribute,
@@ -24,6 +28,8 @@
     "create_glossary_category_assets",
     "create_glossary_assets",
     "create_glossary_term_assets",
+    "list_doc_sources",
+    "fetch_documentation",
     "CertificateStatus",
     "UpdatableAttribute",
     "UpdatableAsset",