You save a hundred articles a year. You write notes in three apps. You bookmark things you intend to read and never see again. The promise of the "second brain" is that all of this scattered knowledge becomes searchable, queryable, and useful. The reality for most tools is that your saved items go into a black hole.
What we actually built
Glassy Second Brain is not a marketing label. It is an indexed knowledge base that contains everything you save — bookmarks, notes, documents, voice transcripts, and your connected Obsidian vault files. Everything is embedded and indexed automatically. BM25 full-text search plus vector semantic search, merged with Reciprocal Rank Fusion. When you search, you get results ranked by relevance across both exact matching and meaning.
Query from Claude, Cursor, Windsurf, OpenCode, Hermes, Openclaw, Pi, or any MCP-compatible client
Here is the part that changes how you use your knowledge. The Glassy MCP server is built on the official
@modelcontextprotocol/sdk v2. It exposes fifteen tools, three prompts, and six dynamic resources.
Generate an MCP key in Settings, paste the config snippet into Claude Desktop, Cursor, Windsurf, OpenCode,
Hermes, Openclaw, Pi, or any MCP-compatible client, and ask questions about your entire corpus.
"What do I know about recursive neural networks?" "Summarize everything I saved about OAuth2 in the last
six months." Your AI assistant queries your actual knowledge base and returns real results with citations.
Privacy by architecture
Your corpus stays on your server. Only the specific query leaves your machine — and that goes to your chosen AI assistant, not to us. The MCP server has DNS-rebinding protection and tier-based rate limiting (120 queries/hr Free, 1,200/hr Pro). It is real infrastructure, not a demo.
How to start
Install the Glassy Companion extension. Save a few pages. Your Second Brain starts indexing immediately. Open Settings → MCP Keys, generate a key, and follow the config snippet for your agent. The whole setup takes under five minutes.