Solo dev's Thea: 1.07M-entry MCP knowledge library for local LLMs, tested vs web search
RespondImpressive172 · reddit · 2026-10-25
A solo Korean developer launched Thea (thealibrary.org), a free knowledge graph that local LLMs query over MCP — "people add the information, AI reads it."
- 1.07M places (mostly Korea, from public data) plus people, works, and organizations; every entry lists sources, author type (human/AI/public/wiki), and verification count. CC BY 4.0.
- Controlled test (same models/prompt, Qwen3-8B & Qwen3.5-35B-A3B, 10 runs each): questions covered by the library score 10/10 in a single call — the 8B got 0/10 with SearXNG web search on the same questions. Out-of-library questions score 0; the model says it doesn't know rather than hallucinating, so it's not a web-search replacement. The 35B could sometimes dig answers from the web but took 3–4x longer. The 8B also answered in Korean after Korean web results; with Thea it answered in English.
- Connects via a Streamable HTTP MCP endpoint (Claude Code, LM Studio, Cursor, or stdio via mcp-remote); tools: theasearch, theafind, theaentry.
The author is soliciting feedback and notes it runs on a free server with rate limits.
More from coding & agent
- Hutash: a two-person team's all-in-one local AI platform now lost in scope — kathi7 · 2026-10-25
- Team ditches Slack for a custom chat app where humans and AI agents work side by side — dee_hw · 2026-10-25
- Musk: Grok Bot can now call businesses for you and summarize the conversation — elonmusk · 2026-10-25
- One prompt let Claude debloat an Android TV over ADB and rebuild a minimal home screen — PrajwalTomar_ · 2026-10-25
- User points Claude at his Samsung TV, spends 11 hours building an ad-free Tizen launcher — PrajwalTomar_ · 2026-10-25
- Jev-Mem paper: a fast scorer makes agent memory building 6.6x faster on LoCoMo — rohanpaul_ai · 2026-10-25