Vellum logo

Vellum

Personal AI Assistant with Persistent Memory and Automation - Vellum

What is Vellum?

Vellum is an AI development platform that lets you build, test, and deploy complex AI workflows and prompt pipelines visually or via code. It helps product and engineering teams experiment faster, evaluate model outputs reliably, and safely launch AI agents into production without managing infrastructure.

Features

Overview

Vellum is an AI assistant platform built around a persistent digital identity rather than a stateless chat window. Each assistant gets its own memory, filesystem, and email address, and it can run background tasks between conversations instead of waiting for a new prompt.

The company originally launched in 2023 as an enterprise LLM development platform, offering prompt engineering, evaluations, and workflow orchestration tools for teams building production AI applications. In May 2026, Vellum pivoted to this personal assistant product, so buyers researching the brand should confirm which version of Vellum a given review or listing refers to.

Vellum runs inside isolated virtual machines and routes tasks across multiple model providers, including Anthropic, OpenAI, Gemini, Kimi, and MiniMax. Users reach it through the web, a native macOS app, iOS and Android mobile apps, Slack, Telegram, email, and the terminal.

The company is Y Combinator-backed (W23) and has raised $25 million in total funding, a $5 million seed round in 2023 followed by a $20 million Series A, with backers including Y Combinator, Rebel Fund, Dharmesh Shah, and Arash Ferdowsi.

Pricing

Vellum’s free Base plan runs forever on a small machine (1 vCPU, 2 GiB RAM) with 6 GiB of storage, pay-as-you-go credits, and Vellum-managed API keys. Paid Pro plans bundle compute, storage, and credits: Mighty is $30/month (1 vCPU, 2 GiB RAM, 10 GB storage, $25 in credits), Super is $100/month (2.5 vCPU, 5 GiB RAM, 30 GB storage, $45 in credits, plus a bundled $10 platform fee and custom email/subdomain), and Ultra is $200/month (4 vCPU, 8 GiB RAM, 60 GB storage, $115 in credits, same platform fee and email/subdomain). A Custom plan lets users configure compute, storage, and credits individually, and a free open-source self-hosted option lets users supply their own API keys and run the assistant on their own hardware with no Vellum platform fees. Background processes such as memory compaction and hourly proactive check-ins consume credits even when the user isn’t actively chatting.

* Disclaimer: Please note that pricing information may not be up to date. For the most accurate and current pricing details, refer to the official website.

Key Features

  • Persistent memory retained across years of conversations

  • Automatic routing across Anthropic, OpenAI, Gemini, Kimi, and MiniMax

  • Native macOS accessibility for desktop and file control

  • Scheduled background routines that run without an open chat

  • Configurable approval levels for sensitive automated actions

  • Bring-your-own-API-key support for self-hosted deployments

Use Cases

01

Automated Inbox Triage

The assistant processes incoming email overnight, sorts messages, archives noise, and drafts replies for the user to approve. This turns an unread inbox into a short list of decisions rather than a wall of messages.

02

GitHub and Linear Triage

Vellum monitors repositories, tags incoming bug reports, and keeps issue boards synchronized with project status. Engineering teams spend less time on backlog upkeep between standups.

03

Meeting Notes and Follow-Ups

The assistant joins calls, captures discussion points, and distributes action items to the relevant channel afterward. Teammates who missed the call get a usable summary instead of a raw transcript.

04

Slack Channel Monitoring

Vellum tracks ongoing Slack discussions, summarizes overnight activity, and flags blockers directly in the channel. It functions as a standing point of reference for teams working across time zones.

05

Personal Scheduling and Travel

Users delegate flight bookings, subscription tracking, and dinner reservations based on stored preferences. The assistant handles the coordination steps that would otherwise require manual follow-up.

Strengths & Weaknesses

Strengths

+

Free open-source self-hosting removes platform fees entirely for users who supply their own API keys.

+

Native macOS accessibility lets the assistant act directly on the desktop and local files.

+

Memory persists across years of use instead of resetting between sessions.

+

Multi-model routing across major providers lets the assistant match models to tasks.

+

Configurable approval thresholds give users control over which actions require sign-off.

Weaknesses

Background routines such as memory compaction and hourly check-ins consume paid credits even without active use.

Desktop computer control is limited to macOS, with Windows support listed only as a roadmap item.

Self-hosted and CLI deployment require comfort managing servers and API keys directly.

The product’s 2026 pivot from an enterprise LLM platform to a personal assistant creates confusion with older reviews of the Vellum name.

Who Is This For?

Founders and business leaders who want a single assistant handling communications, CRM updates, and meeting prep across their existing tools.

Software engineers who want repository triage, GitHub and Linear upkeep, and terminal-based workflows handled automatically.

Busy professionals looking for a personal assistant to manage scheduling, travel, and daily task delegation.

Privacy-conscious users and developers who prefer self-hosting on their own hardware with their own API keys.

Frequently Asked Questions

How does Vellum differ from a standard chatbot like ChatGPT or Claude?

Standard chatbots lose context between sessions, while Vellum keeps a persistent memory, its own filesystem, and an email address, letting it act on tasks without repeated instructions.

Is the free plan actually free, or is it a trial?

The Base plan is free indefinitely, not a time-limited trial. It includes a small machine, 6 GiB of storage, and pay-as-you-go credits.

Why does Vellum cost money even when I’m not actively chatting?

Background systems like hourly check-ins and memory compaction run continuously to keep the assistant proactive, and these consume credits independent of chat activity.

Can I use my own API keys instead of Vellum’s managed plans?

Yes. Vellum supports bring-your-own-key setups through Anthropic, OpenAI, OpenRouter, Fireworks, or custom OpenAI-compatible endpoints, and this is required for self-hosting.

Does Vellum train its models on my data?

No. User workspace data, memories, and credentials are not used for training on Vellum’s side.

Is Vellum the same company that offered an LLM development platform in 2023?

Yes, it is the same YC-backed company, but the product pivoted from an enterprise prompt engineering and evaluations platform to this personal assistant product in May 2026.

What platforms can I use Vellum on?

Vellum is available on the web, native macOS, iOS, and Android, plus Slack, Telegram, email, and the terminal. Windows desktop support is not yet available.

How much storage do the paid plans include?

Mighty includes 10 GB, Super includes 30 GB, and Ultra includes 60 GB of persistent storage, with a Custom plan available for other configurations.

Can I control which actions the assistant is allowed to take on its own?

Yes. Users set an approval level (Strict, Conservative, Relaxed, or Full Access) that determines which file, screen, or email actions require manual confirmation.

What file types can Vellum work with?

It can read and process PDFs, Markdown files, Keynote presentations, spreadsheets, image formats, and zip archives, whether local or uploaded.

Vellum integrates with Gmail, Google Calendar, Slack, GitHub, Linear, Notion, Google Meet, Outlook, HubSpot, Microsoft Excel, Twilio, ElevenLabs, Deepgram, Discord, Telegram, Oura, and Vercel.

Integrations