Vellum

Vellum is a personal AI assistant aimed at professionals who want to offload routine admin work: email triage, calendar wrangling, meeting prep, research, and travel booking. It runs across Mac, Web, iOS, Slack, Telegram, phone, and CLI, and is positioned around a layered memory system (episodic, semantic, procedural, emotional, prospective, behavioral, narrative, shared) that lets the assistant accumulate context about your preferences and working style instead of restarting each session.

The pitch is essentially "ChatGPT with memory, OAuth, and channels" — a managed cloud service with sub-30-second setup, managed OAuth into Gmail/Google Calendar/Slack, and the option to self-host. There's a free tier plus paid plans, and the project is open-source so technically inclined users can run it themselves. It's built on top of frontier LLMs (Claude-class models) rather than a proprietary one.

Closest comparisons are Hermes, OpenClaw, and Claude Cowork. Vellum's edge is the bundled hosting plus OAuth glue, which makes it usable by non-developers; the trade-off is that you're trusting a young product with broad access to your inbox and calendar.