Stats
We may earn commissions from some links.We may earn commissions from links to support our work. Learn more.
Get more AI tool alerts:
About Vellum
Vellum is a personal AI assistant and workspace built around memory, ownership, skills, and always-on execution. It is positioned less like another chatbot tab and more like a personal AI environment: an assistant with its own identity, workspace, tools, channels, and long-term context.
The product’s own language is “the AI layer for your life.” That sounds broad, but the concrete idea is useful: Vellum wants to become the place where an assistant remembers your goals, learns your preferences, connects to your accounts, runs skills, and handles recurring busywork across email, calendar, research, communication, project tracking, travel, and other workflows.
Vellum also has a philosophical wedge: the assistant should be yours. The hosted product gives you a managed cloud assistant with configurable compute and storage. The self-hosted option gives technical users a way to run the open-source assistant on their own hardware, VPS, or private cloud.
If you want to try the assistant, you can start with Vellum here.
What is Vellum?
Vellum is a personal AI assistant platform. It combines chat, memory, workspace files, skills, account connections, scheduling, voice, native channels, cloud hosting, and self-hosting into one assistant environment.
The hosted version runs in Vellum Cloud. The pricing page lets users start free with small processing power and 4 GiB of storage, then upgrade to Pro for configurable compute, more storage, assistant email/subdomain, static IP, and priority support. The self-hosted version is open source and can be deployed on local hardware, a VPS, or private cloud.
That makes Vellum different from a normal AI chat subscription. ChatGPT, Claude, and Gemini are primarily model interfaces. Vellum is trying to be an assistant operating environment: memory plus tools plus permissions plus channels plus long-running autonomy.
Who is Vellum For?
Vellum is for people who want an AI assistant to become part of their operating system, not just answer isolated questions. It will appeal most to users who are willing to invest in setup because they expect the assistant to compound over time.
Specific users who should look at Vellum:
- Founders and operators who want an assistant that can remember goals, prep meetings, triage communication, and handle recurring workflows.
- AI-native professionals who already use agents and want a more persistent assistant layer across devices, accounts, schedules, and tools.
- Technical users who care about self-hosting, open source, data ownership, configurable compute, and transparent usage costs.
- Busy executives or creators who want email, calendar, research, reminders, and drafting handled by a named assistant rather than a generic chat thread.
- Teams evaluating personal AI infrastructure before adopting heavier enterprise knowledge-management systems.
Vellum is probably not for people who just want a cheap chatbot. It is also not for users who dislike granting tool permissions, connecting accounts, or managing AI usage costs. The product’s value comes from integration and persistence, which means trust and setup matter.
Vellum Pros and Cons
| Pros | Cons |
|---|---|
| Memory-first assistant design: Vellum is built around preferences, goals, files, notes, history, and long-term context rather than disposable chats. | More system than casual users need: If you only ask occasional questions, a normal chatbot subscription is simpler. |
| Hosted or self-hosted: Users can choose managed Vellum Cloud or run the open-source assistant on their own infrastructure. | Real cost is subscription plus credits: Pro covers infrastructure; LLM inference, web search, image generation, and paid APIs draw from separate credits. |
| Configurable compute and storage: Pro users can choose machine sizes and storage tiers instead of accepting one fixed SaaS bundle. | Technical responsibility shifts to the user when self-hosted: Deployment, uptime, updates, model/API configuration, and backups become your problem. |
| Assistant can operate across channels and tools: Vellum emphasizes email, calendar, voice, Telegram, Slack, web, macOS/iOS, CLI, and workflow skills. | Trust burden is high: A useful personal assistant needs permissions, account access, memory, and sensitive context. Users need to understand the security model. |
| Transparent pricing philosophy: Vellum says model-provider costs are passed through at cost, with credits used for AI usage and paid third-party APIs. | Still early-category behavior: Personal AI assistants are changing fast; expectations around memory, autonomy, permissions, and reliability are still evolving. |
Vellum’s upside is clear: a persistent assistant that can become more useful every week. Its risk is also clear: the more an assistant can do, the more the user has to trust, configure, and supervise it.
Vellum Features: Memory, Skills, Cloud Hosting & Self-Hosted Control
Personal Memory and Context
Vellum’s main product idea is memory. The assistant is supposed to learn goals, preferences, files, notes, and conversation context over time. That makes it more useful for recurring work than a blank chat window.
The key question is not whether it can remember a single fact. The useful question is whether it can remember enough context to reduce repeated setup: how you like drafts written, what projects matter, what your calendar looks like, which decisions were made last week, and what tasks tend to recur.
Skills and Workflow Automation
Vellum presents skills as adaptive workflows that can handle email triage, calendar changes, meeting prep, delegation, research, communication summaries, project tracking, travel, and other tasks. The homepage examples are ambitious: scanning emails, drafting replies, rescheduling meetings, preparing board-deck summaries, summarizing Slack threads, and booking travel.
The important distinction is that Vellum is not just trying to generate text. It is trying to operate. That makes it closer to an agent runtime than a chat product.
Always-On Cloud Assistant
The hosted version runs in Vellum Cloud. The site emphasizes an always-on assistant, assistant email/subdomain, custom subdomain, static IP, and priority support on Pro. This is useful if you want an assistant that can receive messages, run schedules, and stay available without keeping a laptop awake or maintaining your own server.
That is also why pricing is infrastructure-based. You are not just buying model access. You are buying hosted assistant runtime, storage, and support.
Self-Hosting and Open Source
Vellum’s self-hosting option is a major differentiator. The pricing page says the assistant is open source and can be deployed on local hardware, a VPS, or a private cloud, with no platform fee or machine-tier charges.
Self-hosting is the right path for users who care about control, inspectability, and avoiding lock-in. It is not automatically easier. You still need to manage infrastructure, credentials, providers, backups, updates, and failure modes.
Trust, Permissions and Sensitive Actions
Vellum’s docs emphasize trust: sensitive actions ask permission, users can adjust risk tolerance, and data can live in Vellum Cloud or on a self-hosted machine. The privacy notice also says Vellum may collect conversational content consistent with user preferences for service improvement, but does not use it to train AI models.
That nuance matters. A personal assistant is most valuable when it sees sensitive context. It is also most dangerous when permission boundaries are unclear. Vellum is strongest if it makes those boundaries visible and predictable.
Configurable Compute, Storage and Credits
Vellum Pro separates hosting resources from usage. Base includes Small compute and 4 GiB storage. Pro starts from $50/month using a $10/month platform fee, Medium compute at $35/month, and 10 GiB storage at $5/month. Larger compute and storage tiers cost more, while AI usage is charged through credits.
This is more transparent than a vague all-in subscription, but it means users need to think like operators. Heavy assistant usage can cost more than the base plan.
Vellum vs Alternatives: Pricing & Feature Comparison
| Feature/Aspect | Vellum | Open WebUI | AnythingLLM | Jan | LibreChat | Onyx |
|---|---|---|---|---|---|---|
| Pricing model | Free Base; hosted Pro starts at $50/month plus usage credits; self-hosted option has no Vellum platform fee | Open-source/self-hosted; infrastructure and model/API costs apply | Open-source/local or hosted options; costs depend on deployment and provider usage | Free desktop app; local/API model costs depend on setup | Open-source/self-hosted; infrastructure and API costs apply | Open-source and commercial/team deployment options; infrastructure/provider costs vary |
| Core focus | Personal assistant with memory, skills, channels, cloud hosting, and self-hosting | Self-hosted chat interface over local or API models | Workspace/document chat and local knowledge bases | Privacy-oriented desktop AI chat | Multi-provider ChatGPT-style self-hosted chat with agents and connectors | Enterprise knowledge search with connectors, citations, permissions, and deployment controls |
| Best for | Users who want a persistent assistant that can remember, act, and run workflows | Users who mainly want control over the AI chat UI | People building document/RAG workspaces | Local-first users who want a desktop ChatGPT alternative | Technical teams wanting a self-hosted multi-provider chat app | Teams that need organizational knowledge retrieval, permissions, and enterprise search |
| Memory/assistant depth | Strong personal-memory and assistant positioning | Depends on configured models/tools | Strong for document context, less personal-assistant oriented | More chat-first than workflow-first | Agent/chat oriented; memory depends on setup | Knowledge-search oriented, not a personal assistant |
| Hosting/control | Hosted cloud or self-hosted open source | Self-hosted | Local/self-hosted/cloud depending on setup | Desktop/local | Self-hosted | Self-hosted, cloud, or enterprise deployment |
Vellum should not be compared only to chat apps. The more accurate comparison is personal AI infrastructure. Open WebUI, Jan, and LibreChat are stronger if the job is “give me a controllable chat interface.” AnythingLLM is stronger if the job is “chat with documents and knowledge bases.” Onyx is stronger if the job is enterprise knowledge search. Vellum is most interesting when the job is “give me a personal assistant that remembers, has tools, and can run across channels.”
Vellum Pricing: Plans, Credits & Self-Hosting
Vellum’s public pricing is built from three pieces: the free Base plan, configurable hosted Pro infrastructure, and separate usage credits.
| Option | Price | What it includes |
|---|---|---|
| Base | Free | Small processing power, 1 vCPU / 2 GiB RAM, 4 GiB storage, and pay-as-you-go credits. No credit card required to start. |
| Pro hosted | From $50/month | $10/month platform fee, Medium compute at $35/month, 10 GiB storage at $5/month, assistant email/subdomain, static IP, and priority support. Credits are separate. |
| Self-hosted | No Vellum platform or machine-tier fee | Open-source assistant deployed on local hardware, VPS, or private cloud. Infrastructure, model/API costs, and maintenance still apply. |
Pro can scale beyond the starter configuration. Medium compute is 2.5 vCPU / 5 GiB RAM for +$35/month. Large is 4 vCPU / 8 GiB RAM for +$60/month. XL is 4 vCPU / 16 GiB RAM for +$125/month. Storage tiers range from 10 GiB at +$5/month to 500 GiB at +$120/month.
Credits are separate from the Pro subscription. Vellum states that $1 equals 1 credit. Credits are used for LLM inference, web search, image generation, and paid third-party APIs routed through managed OAuth. Manual top-ups run from $10 to $100 in $10 increments, and auto-reload can add credits when the balance drops below a chosen threshold.
The simplest hosted Pro setup is $50/month before usage credits. That is the number to remember, but it is not the full cost for heavy users.
Is Vellum Worth It? Honest Review
Vellum is worth considering if you want a personal AI assistant that can become infrastructure. The best version of the product is not “another place to chat with Claude.” It is a persistent assistant that knows your work, operates across tools, remembers preferences, and can run in the background.
The hosted Pro plan makes sense if you want that assistant always available without managing your own machine or server. The self-hosted option makes sense if control, ownership, and inspectability matter more than convenience. The free Base plan is the obvious starting point if you want to test the feel of the assistant before paying for larger compute.
The honest limitation is that personal AI assistants are only as valuable as the workflows they actually handle. If Vellum only becomes a nicer chat UI, it will feel expensive compared with a model subscription. If it reliably handles email triage, meeting prep, research, reminders, drafting, summaries, and recurring operations, then the price is easy to justify.
The second limitation is trust. A useful personal assistant needs access. Access creates risk. Vellum’s permission model, privacy posture, and self-hosting option help, but users still need to decide how much autonomy they want to grant.
Vellum Review: Final Thoughts
Vellum is one of the more interesting personal AI assistant products because it is trying to package memory, autonomy, channels, skills, and hosting into one system. It is not the lightest option, and it is not the cheapest way to chat with an LLM. That is not the point.
The point is ownership and persistence. Vellum is compelling if you want an assistant that can accumulate context, operate on your behalf, and exist outside a single browser tab. The hosted plan gives convenience. The self-hosted path gives control. The credit model gives transparency, but it also makes budgeting more important.
I would start with Base, test one real workflow, and then decide. If the assistant saves time on recurring work, Pro is reasonable. If you mostly want a local chat app, choose Jan, Open WebUI, or LibreChat instead. If you want enterprise knowledge search, look at Onyx. If you want a personal assistant that can become part of your life/work operating system, Vellum is worth a serious look.
FAQ
What is Vellum?
Vellum is a personal AI assistant platform with memory, skills, files, conversation history, channels, cloud hosting, configurable compute/storage, and self-hosting.
How much does Vellum cost?
Vellum has a free Base plan. Hosted Pro starts at $50/month before usage credits. The starter Pro configuration is a $10/month platform fee, $35/month Medium compute, and $5/month storage.
Does Vellum Pro include credits?
No. Pro covers hosted assistant infrastructure and Pro features. Credits are purchased separately and spent on LLM inference, web search, image generation, and paid third-party APIs.
Can Vellum be self-hosted?
Yes. Vellum’s assistant is open source and can be deployed on local hardware, a VPS, or private cloud. Self-hosting removes Vellum’s hosted platform and machine-tier fees, but users manage their own infrastructure and AI usage costs.
What are the best Vellum alternatives?
The best alternatives depend on the job. Open WebUI, Jan, and LibreChat are stronger for local or self-hosted chat. AnythingLLM is strong for document/RAG workspaces. Onyx is stronger for enterprise knowledge search. Vellum is more personal-assistant oriented.
