Introduction
Product thesis — why Command Inbox exists and what problem it solves
The problem
For people whose inbox is really a meeting pipeline — founders, recruiters, consultants, agency owners — a large share of email work is scheduling coordination: "does Tuesday work?", "sending an invite", "rescheduling", "following up". Gmail and Google Calendar are two disconnected apps, so this loop often takes ten or more clicks per meeting.
Tools like Superhuman make email faster, but they still treat email and calendar as separate worlds.
The solution
Command Inbox merges Gmail and Google Calendar into one keyboard-driven surface. The core insight:
Every email is either a reply, a meeting, or noise.
The app triages your inbox into exactly those lanes using AI, and makes the email→meeting conversion a single keystroke.
Hero workflow
This is the demo moment judges should look for:
- Open an email thread with scheduling language
- Press
M - AI reads the thread, extracts proposed times and attendees, and shows your calendar availability inline
- Press Enter on a slot
- Calendar invite is sent and an AI-drafted confirmation reply is queued
Roughly three seconds versus two minutes in Gmail + Google Calendar separately.
What makes it different
| Capability | Command Inbox | Typical Gmail client |
|---|---|---|
| Email + calendar in one view | Yes — week strip + thread pane | Separate apps |
| AI triage lanes | Reply / Schedule / FYI / Done | Manual labels or none |
| One-keystroke meeting | M with inline availability | Copy-paste between apps |
| Semantic search over triaged mail | / with pgvector | Gmail search only |
| Agent with Gmail + Calendar tools | Corsair MCP + approval UI | Not available |
| Keyboard-first UX | Superhuman-style shortcuts | Mouse-heavy |
Built for the Corsair hackathon
Command Inbox is a reference implementation of the Corsair SDK stack:
- Gmail + Google Calendar plugins for read/write mail and events
- Webhook-driven cache so inbox reads hit Postgres, not live Google API on every keystroke
- MCP adapter exposing Corsair operations to an in-app agent
- Multi-provider AI (OpenAI + Gemini) for classification, drafts, embeddings, and chat
Next steps
- Evaluators: Read Features and Architecture
- Users: Jump to Getting started
- Developers: Start with Local development