Vellum AI review

★ 4.1/5

Personal AI assistant ("AI chief of staff")

As of mid-2026 vellum.ai hosts "Vellum: Your Personal Intelligence" — an open-source (MIT) personal AI assistant marketed as an "AI chief of staff for high-agency individuals." It lives in Vellum Cloud (or self-hosted), has its own identity/email, a persistent multi-type memory layer, and acts in the world: triages email, manages calendar, posts to Slack/X, updates Linear, makes real phone calls, runs web search, and even vibe-codes small web apps. Capabilities extend via installable "skills" (a catalog at skills.sh incl. Gmail, Slack, DoorDash, Amazon) and via user-added MCP servers (Linear, GitHub, Notion, etc.). NOTE: this is NOT the older Vellum LLMOps/prompt-engineering-and-workflows developer platform that previously occupied this domain; that product appears to have been replaced/repositioned, so any "Pro $500/mo, prompt/workflow execution limits" pricing you find for "Vellum" is stale.

Researched & verified by Dirr Research · scored by our rubric

Visit Vellum AI → last verified 2026-06-02

Affiliate link · how we make money

Best for personal AI assistant. Vellum AI starts Free, then ~$50/mo on a freemium hosting tier (machine + storage) + separate pay-as-you-go credits ($1 = 1 credit) for LLM/tools model. The Pro plan's headline "$50/mo" is purely hosting: $10 platform fee + $35 Medium machine + $5 for 10 GiB storage, and it includes ZERO AI usage. Add realistic usage of ~$40/mo in credits (LLM tokens, web search, phone calls, billed at $1 = 1 credit with no markup) and a real Pro user lands around $50 + $40 = ~$90/mo. Self-hosting the open-source build with your own API keys drops the $50 hosting layer to $0, leaving only your model/provider costs.

Spec

pricing modelfreemium hosting tier (machine + storage) + separate pay-as-you-go credits ($1 = 1 credit) for LLM/tools
entry priceFree, then ~$50/moverify livedecode →
free tierYes - Base plan free forever, no card: pay-as-you-go credits, fixed Small machine, 4 GiB storagesource →
self-hostYes
MCPClient
integrationsChannels: Telegram, Slack, Email, Phone (voice calls); Clients: macOS, iOS, Web, CLI, Chrome extension. Skills catalog (Gmail, Google Calendar, Slack, Amazon, DoorDash, etc.) plus any MCP server (Linear, GitHub, Notion, Context7) added via CLI.source →
API accessYes
affiliate30% of a referred customer's first payment

The gotcha

the pricing catch

The Pro "$50/mo" buys only the hosting shell (machine size, storage, custom subdomain, static IP, priority support) and includes NO credits — every bit of actual AI work (LLM inference, web search, image gen, phone/3rd-party APIs) is billed separately as pay-as-you-go credits on BOTH the Free and Pro tiers, so your real bill is hosting + whatever your assistant consumes.

Decode the real cost →

Best for

  • You are a founder/operator who wants an always-on, memory-rich AI assistant that owns busywork (inbox, calendar, Slack, calls) across Mac/iOS/web/Telegram, and you are comfortable paying model costs as pay-as-you-go credits — or self-hosting the open-source version on your own keys.

source: vellum.ai/pricing →

Vellum AI pricing tiers

Prices move. Cells flagged verify link to the live vendor page.

Decode Vellum AI pricing →
tiermonthlyannualincludedunit
Base $0 n/a Free forever, no card. Pay-as-you-go credits, fixed Small machine, 4 GiB storage. No custom subdomain/email. per month
Pro $50verify n/a From $50/mo = $10 platform fee + Medium machine ($35) + 10 GiB storage ($5). Configurable machine/storage, assistant email + custom subdomain, static IP, priority support. Credits NOT included. per month (base config)
Machine add-on (Pro) $35verify n/a Medium 2.5 vCPU/5 GiB +$35; Large 4 vCPU/8 GiB +$60; XL 4 vCPU/16 GiB +$125 per month per machine size
Storage add-on (Pro) $5verify n/a 10 GiB +$5; 30 GiB +$10; 60 GiB +$15; 120 GiB +$30; 250 GiB +$60; 500 GiB +$120 per month per tier
Credits $10verify n/a Pay-as-you-go, $1 = 1 credit, bought in $10 increments up to $100/top-up; consumed by LLM inference, web search, image gen, paid 3rd-party APIs. No markup on token usage. per credit ($1)
Self-hosted (open source) $0verify n/a MIT-licensed; deploy on own hardware/VPS/private cloud. No platform fee, no machine charges; you supply your own keys/uptime. free (BYO infra + model costs)

// pools and per-unit rates are volatile · cells flagged verify link to the live vendor page

Compare Vellum AI with…

vellum ai pricing decoded→ vellum ai alternatives→

Questions

How much does Vellum AI cost?

Vellum AI starts Free, then from ~$50/mo. The Pro plan's headline "$50/mo" is purely hosting: $10 platform fee + $35 Medium machine + $5 for 10 GiB storage, and it includes ZERO AI usage. Add realistic usage of ~$40/mo in credits (LLM tokens, web search, phone calls, billed at $1 = 1 credit with no markup) and a real Pro user lands around $50 + $40 = ~$90/mo. Self-hosting the open-source build with your own API keys drops the $50 hosting layer to $0, leaving only your model/provider costs.

Can you self-host Vellum AI?

Yes, Vellum AI can be self-hosted.

Does Vellum AI support MCP?

Vellum AI supports MCP as Client, so it works with assistants like Claude and ChatGPT.

What's the catch with Vellum AI's pricing?

The Pro "$50/mo" buys only the hosting shell (machine size, storage, custom subdomain, static IP, priority support) and includes NO credits — every bit of actual AI work (LLM inference, web search, image gen, phone/3rd-party APIs) is billed separately as pay-as-you-go credits on BOTH the Free and Pro tiers, so your real bill is hosting + whatever your assistant consumes.