"Organize your conversation logs with terminal agents (Claude Code, Codex, Gemini CLI) in Obsidian, use them as memory, and turn them into a wiki - the current desktop app supports up to Claude export .zip files"
I built it using the session search app made by someone famous(?) as QMD and a wiki document I saw on Hacker News (https://gist.github.com/karpathy/442a6bf555914893e9891c11519de94f).
It's "an app that turns the conversations you had with terminal agents (Claude Code, Codex, Gemini CLI) into a wiki format in Obsidian and makes them easy to search."
If you apply it, you can give terminal agents something similar to long-term memory (not quite, more like searching old conversations). (e.g., How did I patch that upstream error from last time? If it's in Obsidian, it'll find and tell you - if not, into the abyss of hallucination...)
These days, since I do almost 80% of everything by talking with and building things alongside terminal agents, I thought it'd be nice to organize the chats a bit. Then one day while googling, I found that someone had made something similar. They were a foreigner, but at the time the search algorithm (bm25) didn't work well in Korean, so it seemed like it would take a lot of effort, and I just gave up partway. But I had some Claude usage left over, so I somehow solved(?) the Korean search problem and built it. I feel like I talked with the agents quite a lot, but only a few wiki entries came out lol. Still, it's nice to have something to read. If you use terminal agents, just clone it from GitHub, ask them to install the repo and run it, and they'll do it for you. (Honestly, I always do it that way too...) I haven't released it yet, but I also built something called tunaflow, a multi-agent orchestration app, partly as a test. There's a delay of a few seconds in responses due to a terminal subagent tool loading issue, but the performance is solid so far lol (actually it might just be Opus's ability ㅠ).
**If you register it as an mcp in Claude Code, you can search previous conversations within the integrated dialogue!**
Anyway, something similar has already been implemented, but I just advanced Korean search a bit. When I asked the AI to summarize what it is, it described it grandly like below.
If you use terminal agents a lot, it might be worth giving organizing a try! (Honestly, it's not as grand as that?)
Of course it's open source, and I'd be so grateful if you'd give it a star, and feel free to fork it and chew, bite, and enjoy it however you like 😁
"seCall is a local-first search engine for AI agent sessions. It collects conversation logs from Claude Code, Codex CLI, and Gemini CLI, indexes them with BM25 + vector hybrid search, and provides them as a CLI/MCP server/Obsidian-compatible knowledge vault"
https://github.com/hang-in/seCall
