IntoAction

Most CRMs die because nobody fills them in. IntoAction skips the data entry: it ingests the communication you already have, email, calendar, messaging, and builds the relationship picture itself, then prompts action before a contact goes quiet for six months.

What it does#

• Ingests broadly. Around 25 provider adapters spanning email, calendars, meeting tools, chat platforms and the big CRMs, feeding one pipeline.
• Keeps the picture current. The AI layer turns raw communication into an up-to-date view of every relationship, personal and commercial, without a human maintaining it.
• Prompts the follow-up. “I should reply to that” becomes an edit rather than a chore, across web and mobile.

The interesting problem#

The adapter layer is the real work. Twenty-five providers means twenty-five opinions about auth, pagination, rate limits and what a “message” even is, and the ingestion pipeline has to normalise all of it into one interaction model. The database has the same split personality by design: one Drizzle schema that runs on SQLite for local-first use and PostgreSQL in production, which forces honesty about what the data layer actually depends on.

The stack#

A TypeScript monorepo with pnpm and Turborepo: a Hono API, a Next.js 14 web app, and an Expo React Native mobile app, with Drizzle ORM over the dual-dialect database and Anthropic plus OpenAI for the AI layer.