---
source_url: "https://www.vellum.ai/llm-leaderboard/ai-assistants/vellum"
title: "Vellum AI Assistant Review 2026 - Features, Pros & Cons"
mirrored_at: 2026-08-10T01:04:34.811Z
host: www.vellum.ai
cited_in_42a: true
mirror_canonical: "https://index.42a.ai/www.vellum.ai/llm-leaderboard/ai-assistants/vellum"
---

> **Original source:** https://www.vellum.ai/llm-leaderboard/ai-assistants/vellum

Your personal AI assistant that remembers, learns, and takes real action across every platform you use.

[www.vellum.ai](https://www.vellum.ai/)

Vellum is a personal intelligence platform with a persistent AI assistant that has deep memory, its own identity, and the ability to act autonomously. It can send emails, manage calendars, browse the web, control your Mac, and more.

It offers one-click cloud deployment or fully self-hostable infrastructure. The three-layer memory system (workspace files, curated personal knowledge base, and auto-extracted long-term memory with confidence scoring and reinforcement tracking) sets it apart from every other assistant.

## Overview

License

MIT

Pricing

Free + API costsPaid plans available for managed cloud

Ease of setup

Very easyOne-click cloud deploy or self-host

Cloud hosting

Managed cloudAlways-on sandboxed environment

Self-hosted

YesDesktop app, ~/.vellum/workspace/

Multi-model

Any LLMOpenAI, Anthropic, Google, OpenRouter, Fireworks, Ollama

## Strengths & Weaknesses

Strengths

Easiest setup. One-click cloud deploy, no Docker or config files

Broadest channel coverage of any assistant (9 channels)

Best-in-class persistent memory with three layers

Enterprise-grade security with sandboxed credentials

Managed OAuth for 10+ services

Multi-model support including open-source LLMs

Self-improving skills that learn from interactions

Weaknesses

No Windows app yet

Plugin ecosystem still growing compared to OpenClaw/Hermes

Paid plans required for full managed cloud features

## Links

[Website](https://www.vellum.ai/)[Documentation](https://www.vellum.ai/docs)[llms.txt](https://www.vellum.ai/llms.txt)[Skills Catalog](https://skills.sh/)[Compare all assistants](https://www.vellum.ai/llm-leaderboard/ai-assistants)

## Capabilities

Memory & Intelligence

Persistent memory

Managed memoryThree-layer system with auto-extraction, confidence scoring, reinforcement tracking

Self-improving skills

YesLearns patterns, habits, preferences from interactions

Cross-session context

Full context carryoverIDENTITY.md, SOUL.md, USER.md, NOW.md loaded every session

Security

Credential management

Sandboxed credentialsIsolated per-pod credential executor; LLM never sees raw values

Permission system

Custom permissioningGranular tool-level access controls, risk tolerance tiers

Network isolation

Pod-level isolationEach assistant runs in its own Kubernetes pod

Integrations

OAuth connections

Managed OAuthGoogle, GitHub, Notion, Linear, Twitter/X, Outlook, and more

MCP support

Yes

Plugin ecosystem

GrowingSkills catalog at skills.sh; custom skills via SKILL.md

Automation

Scheduled tasks

YesRRULE, ISO 8601, execute and notify modes, timezone-aware

Heartbeat monitoring

Yes

## Channels & Platforms

iOS

Yes

Native app

macOS

Yes

Native app

Windows

No

Web

Yes

CLI

Yes

Voice

Yes

Built-in via Twilio

Email

Yes

Own email address

Telegram

Yes

Slack

Yes

Socket Mode

## Related Articles

[Vellum vs OpenClaw](https://www.vellum.ai/compare-assistants/vellum-vs-openclaw)[Vellum vs Claude Cowork](https://www.vellum.ai/compare-assistants/vellum-vs-claude-cowork)[Vellum vs Hermes Agent](https://www.vellum.ai/compare-assistants/vellum-vs-hermes-agent)[Vellum vs Manus](https://www.vellum.ai/compare-assistants/vellum-vs-manus)[Best Personal AI Assistants 2026](https://www.vellum.ai/blog/best-personal-ai-assistants-2026)[Best Personal AI Assistants for Developers](https://www.vellum.ai/blog/best-personal-ai-assistants-for-developers)[Best Personal AI Assistants with Memory](https://www.vellum.ai/blog/best-personal-ai-assistants-with-memory)