The Context Layer

The Sense → Decide → Act → Close loop

Everything Lodestar does — and everything an AI connected via MCP can do — follows the same four-step loop.

• Sense: ingest raw signals from email, calendar, files, and messages
• Decide: the AI (Claude, ChatGPT, or Lodestar's own assistant) reasons over the enriched context
• Act: write back — create a follow-up, draft a reply, update a milestone, log an event
• Close: log the action, notify stakeholders, and mark the loop complete

How Lodestar enriches your data

Raw email, calendar events, and files are not directly useful to an AI — they're too noisy and too large. Lodestar processes them into structured intelligence: action items with evidence quotes, contact sentiment scores, org chart relationships, project-linked timelines, and a ranked priority feed.

This enriched layer is what the MCP server exposes to your AI. The AI gets the signal, not the noise.

Model-agnostic by design

Lodestar does not lock you to one AI model. The MCP server works with any client that speaks the MCP protocol. When a better model ships, your accumulated context becomes more valuable — you don't start over.

Your data stays yours

All data is encrypted per user. Lodestar never trains on your data. You can bring your own model API key (BYOK) so inference runs through your own account. You can export or delete everything at any time.