Vellum
A personal AI assistant that actually remembers things between conversations — your preferences, ongoing projects, past decisions — and can take real actions like managing your calendar or email instead of just chatting and forgetting everything the momen
🔗 Visit Vellum
Description
Most AI chatbots start every conversation from zero: you re-explain context, re-paste the same background information, and nothing you told it yesterday carries over to today. Vellum is built around the opposite idea — a persistent memory that accumulates your preferences, files, notes, and conversation history over time, so it behaves more like an assistant who's worked with you for months than a stranger you're meeting each session.
Beyond memory, it can browse the web, manage files, execute code, and handle calendar and email tasks, with integrations into Slack, Telegram, and GitHub. It runs background and scheduled tasks (things it does on its own timeline, not just when you ask), is deployable across web, macOS, iOS, CLI, and even email, and is designed to improve its own workflows over time. Pricing includes a free tier, a Pro tier at $50/month base plus compute add-ons ($35-125/month) and storage add-ons ($5-120/month), with credits billed at $1 each for LLM inference, web search, image generation, and third-party API calls. Your workspace is explicitly user-owned, with data export and deletion available.
💬 Our review
The short version: Vellum is worth trying if the thing that's frustrated you about AI assistants is having to re-explain yourself every session — persistent memory across conversations is the core differentiator here, and it's a genuinely different product shape than a stateless chatbot.
The pricing structure is the thing to study closely before committing: a $50/month base fee plus separate compute add-ons ($35-125/month) and storage add-ons ($5-120/month), on top of $1-per-credit billing for actual usage (inference, web search, image generation, third-party APIs), means your real monthly cost could range widely depending on how heavily you use it — budget for the higher end rather than assuming the base fee is the whole bill. Against a general-purpose assistant like ChatGPT with memory features, or Claude with its own project-context tools, Vellum's differentiator is being positioned as a dedicated 'personal AI environment' with its own identity, scheduled background tasks, and deep integrations (Slack, Telegram, GitHub) rather than a chat interface you occasionally ask things of. Given it's marketed as user-owned data with export and deletion built in, that's a reasonable trust signal, but as with any assistant that has calendar/email access and runs tasks on its own schedule, it's worth starting with lower-stakes tasks before handing it anything sensitive. <!-- ai-generated -->
💰 Pricing
📊 Global score
🤖 AI-enriched data
Gratuit disponible ; Pro 50$/mois de base + compute 35-125$/mois + stockage 5-120$/mois ; crédits à 1$ chacun (inférence LLM, recherche web, génération d'image, API tierces)
Pros
Mémoire persistante entre les conversations, contrairement à un chatbot classique qui repart de zéro
Peut exécuter des tâches en arrière-plan et planifiées, pas seulement répondre sur demande
Déployable sur web, macOS, iOS, CLI et même email
Données explicitement possédées par l'utilisateur, export et suppression disponibles
Cons
Structure de prix complexe (base + compute + stockage + crédits) — coût réel difficile à anticiper
Accès calendrier/email à un agent autonome = prudence recommandée sur les tâches sensibles au début
Pas de liste claire de concurrents directs communiquée par l'entreprise
Produit encore jeune, peu de recul sur la fiabilité de la mémoire à très long terme