How It All Fits Together
Vellum is a personal assistant that evolves with you — it learns how you think, work, and communicate, and gets better every conversation.
Here's how the pieces connect.
What happens when you send a message
- You send a message through any channel — the desktop app, Telegram, voice, Slack, or others.
- The assistant gathers context — your identity and preferences, relevant memories from past conversations, and available skills.
- An AI model reasons about your request and decides what to do — it may read files, send emails, browse the web, control your screen, or take other actions.
- Results flow back through the same channel you started from.
- The assistant remembers what it learned — facts, preferences, and context — for future conversations.
Computer use
On macOS, the assistant can see and control your screen. This is one of its most distinctive capabilities.
| Step | What happens |
|---|---|
| Perceive | Reads your screen two ways — the accessibility tree (the same API screen readers use) and screenshots — for both structured and visual understanding. |
| Verify | Every action passes through safety checks: sensitive data detection, destructive action blocking, loop detection, and step limits. |
| Execute | Injects mouse clicks and keyboard input via macOS system events. Text input uses the clipboard to handle special characters correctly. |
| Observe | Waits for the UI to settle, captures the new state, and decides what to do next. |
This loop runs for up to 50 steps per session. You see an overlay showing what the assistant is doing, and you can stop it at any time.
Memory
The assistant remembers what matters. As you talk to it, it picks up facts, preferences, and context from your conversations and stores them automatically.
When you ask something later — even in a different conversation — it doesn't just match keywords. It understands the intent behind your question and finds relevant memories even when you phrase things differently from how they were originally stored. Results are diversity-ranked so you get a broad picture, not five variations of the same fact.
The assistant also keeps a journal — a running narrative of what's happening in your life and work. Important entries carry forward automatically, while routine ones naturally fade over time. Occasionally, it surfaces older memories you didn't search for, so useful context doesn't get buried.
The result: your assistant gets better the more you use it, without you having to explicitly teach it anything.
Trusted contacts
You can grant other people limited access to your assistant through a guardian system.
A trusted contact can be verified through external channels like Telegram, Slack, or phone. Once verified, they can send messages to your assistant and, depending on the permissions you've set, approve sensitive actions on your behalf.
This is useful for delegation: your assistant can handle a request from a colleague, check with you before doing anything risky, and report back through whatever channel they're on.
Skills in depth
The assistant ships with dozens of bundled skills, with more available from the community:
| Category | Examples |
|---|---|
| Communication | Gmail, Slack, phone calls, messaging, email sequences |
| Research & browsing | Browser automation, computer use |
| Productivity | Contacts, tasks, followups, notifications, playbooks, Google Calendar, scheduling |
| Monitoring | Screen watch (OCR-based), watcher (polling external sources), heartbeat |
| Content & media | Long-form writing, image generation, media processing, transcription |
| Development | App builder, subagent orchestration, ACP (external coding agents), skill management |
Each skill is a self-contained package: instructions that teach the assistant when and how to use it, a tool manifest defining what actions are available, and the implementation code behind those tools.
You can also create your own skills or install community-built ones. Describe what you want, and the assistant will scaffold, test, and persist a new skill to your workspace.