Open-Source 'The Librarian' MCP Server Manages 3,000 Skills, Saves Tokens and Self-Optimizes
Open-Appeal-9747 · reddit · 2026-08-04
Reddit user Open-Appeal-9747 shares an open-source MCP server called The Librarian that addresses token consumption and skill selection in AI agent skill management.
Problem: Every installed skill's description is loaded into agent context on each request, causing high token costs and overlapping descriptions leading to wrong skill triggers.
Solution:
- Skills stay as raw markdown on disk, never entering context.
- Agents describe tasks in plain language; Librarian recommends the few best-fitting skills via local embedding search with anti-repetition.
- Agents report whether skills worked; unused skills become kill candidates, unmatched queries mark gaps, enabling self-curation.
Technical details:
- Fully local: Ollama embeddings + SQLite, optional Apple on-device reranking.
- Python single file, MIT license, requires Python 3.12+ and Ollama.
- Includes real usage data: 130 queries, 76 outcome reports.
GitHub: https://github.com/aka-kika/the-librarian
More from coding & agent
- BitBox: Non-engineers successfully audit codebase using AI agents — RSync25 · 2026-08-04
- Is Traditional RAG Dead? Developers Discuss Six Alternative Architectures — SwordfishWest6860 · 2026-08-04
- Alibaba Amap Open-Sources LongHorizon-Harness: Boosting Agent Task Success Rates via Auditing — aigclink · 2026-08-04
- Indie Dev Builds Universal Memory Layer Across Models: No Vendor Lock-in, Predictable Cost — Opening-Dream9276 · 2026-08-04
- Pro Tip: Ask Claude to Generate Architecture Diagrams for Large Codebases — aniketmaurya · 2026-08-04
- Swapping Codex to Kimi K3: 1/4 the Cost of GPT-5.6 for Marketing Plans, Minor Quality Drop — AskItAll_Biych · 2026-08-04