--- layout: trip2g/instaframe free: true --- Draft deck 2: the MCP knowledge base. An agent searches, descends, cites real passages. ### NOTES AN AI CAN [WALK] ### CHATBOTS DON'T KNOW [YOU] Ask a chatbot about your own research and it improvises. It never read your notes. It's guessing from the internet's average opinion. ### PASTING NOTES DOESN'T [SCALE] Copy a note into the chat, get one decent answer, repeat forever. Your vault has a structure. The paste buffer doesn't. ### GIVE THE MODEL A [MAP] trip2g exposes your vault to AI agents over MCP, the protocol tools like Claude speak. The agent doesn't get a text dump. It gets search, links, and structure. ### IT SEARCHES, THEN [DESCENDS] The agent queries your base, opens the most promising note, follows its wikilinks deeper. The same moves you'd make. Just faster, and it never gets bored. ### IT CITES REAL [PASSAGES] Answers come back with quotes pulled from your actual notes. Not "studies suggest". Your words, your sources, traceable to the exact note. ### WATCH IT [THINK] There's a search visualizer: you can watch the model walk your graph, hop by hop. Which note it opened, which link it followed, where it found the answer. ### YOUR LINKS FINALLY PAY [OFF] Every wikilink you ever made is now a road an agent can travel. Ten years of linking notes turns out to be ten years of building infrastructure. ### GARBAGE IN, GARBAGE [OUT] The agent is only as good as the base it walks. Which is the point: it rewards you for keeping notes clear and connected. ### WHAT PEOPLE [BUILD] A book-notes base you can interrogate. A team wiki that answers questions. A research corpus where "where did I read that?" takes five seconds. ### TRY IT ON YOUR [VAULT] trip2g is open source. Point it at your markdown, connect an agent, ask something hard. Then watch it walk. ### SAVE THIS [IDEA] Your second brain can finally talk back. Save this for when you're ready to introduce it to an AI.

notes-an-ai-can-walk