Docs / Getting Started / What is Vellum?

What is Vellum?

The short version

Vellum is a personal AI assistant that lives on your computer. It has its own identity, its own email, its own accounts. It can read your files, manage your calendar, order you food, build you an app, and remember that you take your coffee black.

It's not a chatbot. It's not an autocomplete engine. It's a separate entity that works for you, learns about you, and takes real actions in the real world.

How is this different from ChatGPT / Claude / etc.?

Those are conversation tools. You type, they respond, you copy-paste the answer somewhere useful. The conversation ends and everything resets.

Vellum is different in a few important ways:

It has tools, not just words. Your assistant can browse the web, read your files, run code, send emails, manage your calendar, and interact with dozens of services. It doesn't describe what you could do. It does it.

It remembers you. Not just within a single conversation. Across days, weeks, months. Your preferences, your projects, your quirks. It builds a picture of who you are and uses that to help you better over time.

It has its own identity. Your assistant isn't borrowing your email or your GitHub account. It has its own. When it sends an email on your behalf, the recipient knows they're talking to your assistant, not to you. Clear boundaries, no confusion.

It lives on your machine. Your workspace, your memories, your configuration... all stored locally as plain text files you can read and edit. No mysterious cloud database. No data you can't access. It's yours.

It's personal. Not a team tool. Not a shared resource. Not a Slack bot everyone in your company uses. It's your assistant, personalized to you, and nobody else can access it.

What can it do?

A non-exhaustive list of things your assistant can handle:

  • 📧 Email — Read, search, draft, and send emails from its own address
  • 📅 Calendar — Check your schedule, create events, find availability
  • 🌤️ Weather — Current conditions and forecasts, any location
  • 🛒 Food delivery — Order from DoorDash (yes, really)
  • 🖼️ Image generation — Create and edit images with AI
  • 🌐 Web browsing — Search the internet, read pages, extract information
  • 💻 Coding — Write, run, and debug code in a sandboxed environment
  • 📱 Messaging — Interact across Slack, Gmail, Telegram, and more
  • 🏗️ App building — Create interactive HTML/CSS/JS apps on the fly
  • 📝 Writing — Draft documents, blog posts, reports in a full editor
  • Tasks & reminders — Manage your to-do list, set schedules, track follow-ups
  • 🔍 Research — Search, synthesize, and present findings on any topic

And if it can't do something? You can teach it. Skills are modular and extensible. More on that in Your First Skill.

How does it work? (The 30-second version)

  1. You talk to it. Through the desktop app, voice chat, or (eventually) other channels.
  2. It thinks. Your message, along with relevant context (your preferences, memories, workspace files), is sent to a cloud AI model to generate a response.
  3. It acts. If the response involves doing something (reading a file, sending an email, building an app), your assistant uses its tools to make it happen. Sensitive actions require your explicit permission first.
  4. It learns. Important facts, preferences, and context are saved to your local workspace so it gets better over time.
Transparency moment: Your assistant runs locally, but it thinks in the cloud. Your prompts and context are sent to an AI model provider (like Anthropic) to generate responses. This is the trade-off of having a smart local assistant. We'd rather tell you upfront than have you discover it in a footnote.