Most AI assistants re-read your inbox and transcripts cold on every request, so the context dies the moment the chat window closes. The bet here is different: the real bottleneck isn't model quality, it's whether the assistant keeps a durable, inspectable memory of your work — and that memory should be plain files you own, not a vendor's black box.
What Sets It Apart
- Turns Gmail, Calendar, and meeting notes (e.g. Fireflies) into a knowledge graph stored as an Obsidian-compatible vault of Markdown with backlinks — you can open, edit, and grep it directly.
- "Live notes" auto-update around people, topics, and projects as new context arrives, so briefs and recaps stay current without manual stitching.
- Acts on that graph: generates email drafts, briefs, decks, and PDFs grounded in your accumulated context, with voice in/out via Deepgram and ElevenLabs.
- Runs against local models (Ollama, LM Studio) or hosted APIs, and extends to external tools through MCP — no single-provider lock-in.
Great Fit If
You live in email, meetings, and notes and want an assistant whose memory compounds over weeks rather than resetting each chat — and you value owning your data as plain Markdown you can inspect. Look elsewhere if you want a turnkey hosted product: this is self-hosted, wires into your own accounts and model keys, and the graph's usefulness depends on feeding it your real workflow.
