MCP Overview

What the Lodestar MCP server exposes and how AI assistants use it to read context and take action.

What the MCP server does

The Lodestar MCP server is the bridge between your enriched work context and any MCP-compatible AI assistant. It provides two categories of capability: read tools (pull your dashboard, emails, files, calendar events, sentiment data, and knowledge graph) and action tools (create follow-ups, draft reply links, update action items, log timeline events).

Connect it once and every conversation with Claude, ChatGPT, or Claude Code can start from your actual operating picture.

  • Read: dashboard_today, needs_attention, sentiment, emails, files, calendar, knowledge_search
  • Act: action_items_create, action_items_update, emails_reply_link, timeline_events_create
  • Model-agnostic — works with any MCP-compatible client

Sense → Decide → Act → Close

Every useful AI interaction over MCP follows the same pattern: the AI reads context (Sense), reasons over it (Decide), takes an action (Act), then logs the result (Close). The MCP server provides the tools for each step.

text
SENSE  ──▶  DECIDE  ──▶  ACT  ──▶  CLOSE
(read       (the AI      (write      (log + notify
 context)    reasons)     back)       + track)

Prerelease availability

The MCP server is available to prerelease members. The endpoint URL and full tool reference will be finalised at GA. Connection steps in the following pages use a placeholder URL.

To get MCP access, join early access at getlodestar.ai/early-access. Your MCP credentials will be provided during onboarding.