Platform

Knowledge and memory

The layer that makes an answer specific: your documents, your history, and what Lodestar has learned about how you work.

The knowledge graph

People, companies, projects and topics, connected by what actually passed between them. Explore it as a graph or a structured tree, change how far out it reaches, and open any node to see the mail, meetings and files behind it.

Answers grounded in your own material

Add reference documents to a knowledge base and Lodestar indexes them. On every question, relevant passages are retrieved and put in front of the model before it answers — not fetched afterwards to justify a guess — and the sources are listed so you can check them.

Memory that persists

Preferences and durable facts carry across conversations. You can tell it to remember something, tell it to forget, and see and edit the whole list in Settings. Nothing is remembered that you cannot inspect.

You can see what it is working from

The assistant shows a breakdown of everything filling its context on any turn — instructions, tools, retrieved passages, conversation — and what the turn cost. Most tools ask you to trust the answer; this one shows the inputs.

  • ·Interactive knowledge graph over people, companies and projects
  • ·Retrieval over your documents, mail, files and messages
  • ·Editable long-term memory
  • ·Per-turn context and cost breakdown