This turns a bunch of links into a robust library that can be queried with an mcp server.
MCP Librarian: The Knowledge Indexer (mcp-link-library)
The Universal Search & Retrieval Engine for the Workforce Nexus.
The Librarian (mcp.py) is a specialized MCP server responsible for indexing, storing, and retrieving knowledge from heterogeneous sources. In v3.3.1, it includes the ATP Sandbox for secure execution of agent-driven logic.
🚀 Quick Start (Suite Mode)
The Librarian is automatically registered and monitored by the Dashboard:
../nexus.sh
Standalone Start:
python3 mcp.py --server
🌟 Capabilities (v3.3.1)
1. Universal Indexing (v3.0)
Deep content parsing for Documents (PDF, DOCX, XLSX) and Images (EXIF).
- Optimization: Server-side extraction returns only relevant JSON paths, reducing token bloat by 80%.
2. ATP Sandbox & Security
Includes the ATPSandbox layer to protect the host environment:
- AST Guards: Blocks
type(),getattr(), and class-escape vectors. - Portability: Uses deterministic wrappers to prevent infrastructure hallucinations.
3. Self-Healing & Diagnosis
check_health: Audits Python dependencies and database integrity.update_dependencies: One-click recovery via the Nexus Dashboard.
4. Zero-Token Filtering
- Paginated Lists: capped at 50 results to prevent context saturation.
- Deep Search: Returns snippets instead of full documents for efficiency.
🛠️ Global Command Reference
| Command | Action |
| :--- | :--- |
| mcp-librarian --search X | Search the knowledge base for X. |
| mcp-librarian --add Y | Index a new URL or local file. |
| mcp-librarian --list | List all indexed resources (summarized). |
📝 Metadata
- Status: Production Ready (v3.3.1)
- Author: l00p3rl00p
- Part of: The Workforce Nexus Suite