Vellum

Vellum is a personal AI assistant that can read files, manage calendars, send email, run code, and control a computer. It saves preferences and context in its workspace for later conversations, while custom skills let users extend its capabilities. Listed OAuth services include Discord, GitHub, Google services, HubSpot, Linear, Twitter (X), Asana, Notion, Microsoft Outlook, and Todoist. Vellum says OAuth tokens are kept in a credential vault and are not exposed to the language model. Sensitive actions prompt for approval, and users can set trust rules for ongoing permissions. By default, conversation context is sent to Anthropic to generate responses; OpenAI, Google, OpenRouter, Fireworks, and local Ollama are listed as alternatives. An internet connection is required for responses. Vellum Cloud runs assistants in isolated environments with private storage and 24/7 availability. The Free plan includes initial tasks and pay-as-you-go credits; Mighty costs $30.00 per month. There is no shipped Linux desktop client, though Linux users can use the web app, Chrome extension, or self-host the runtime.

Vellum combines personal task assistance with saved context, custom skills, and approval controls. Its Free plan includes initial tasks and pay-as-you-go credits; paid plans start at $30.00 per month.

Vellum suits individuals looking for an AI assistant to handle personal tasks across files, calendars, email, code, and computer controls. It is described as a personal assistant rather than a shared company tool.

Overview

Vellum is a personal AI assistant for people who want help carrying out tasks across files, calendars, email, code, and a computer. It is best suited to individual users willing to configure skills and permissions; its broad reach comes with a firm requirement for internet access.

Saved context, OAuth connections, and approval controls make Vellum more than a prompt-and-response tool. It is a poor fit for teams seeking a shared company assistant, or for anyone who needs an assistant to work offline.

Key features

Vellum saves preferences and context in its workspace for use across conversations. That continuity can spare users from repeating background, while custom skills that users create, install, and manage let them extend what the assistant can do. The trade-off is that the experience is geared toward people prepared to tailor an assistant rather than simply use a fixed set of functions.

OAuth connections cover Discord, GitHub, Google services, HubSpot, Linear, Twitter (X), Asana, Notion, Microsoft Outlook, and Todoist. Vellum says tokens remain in a credential vault and are not exposed to the language model. Sensitive actions prompt for approval, and configurable trust rules allow standing permissions. Those controls matter for an assistant that can act across connected services, though users still need to decide which actions warrant ongoing trust.

Vellum Cloud runs assistants in isolated environments with private storage and 24/7 availability. Vellum says it encrypts data end-to-end, conducts continuous security audits, and does not use user data for model training. Optional usage analytics and crash diagnostics are off by default. Conversation context goes to Anthropic by default; OpenAI, Google, OpenRouter, Fireworks, and local Ollama are also model options. Even with local Ollama, Vellum says an internet connection is required to reach an AI model provider, so it is not an offline assistant.

Pricing

The Free plan costs nothing and includes initial tasks plus pay-as-you-go credits. It offers a way to start without a subscription, but ongoing use depends on credits rather than an included usage allowance.

Mighty costs $30 a month, billed $30/month. It includes a small computer with 1 vCPU and 2 GiB, 10 GB of storage, and Mighty Usage. This is the entry paid tier for an individual who wants a hosted assistant with included usage.

Super costs $100 a month, billed $100/month. It steps up to a medium computer with 2.5 vCPU and 5 GiB, 30 GB of storage, Super Usage, and assistant email and subdomain. It suits users who need more hosted resources or want those assistant-specific contact details.

Ultra costs $200 a month, billed $200/month. It includes a large computer with 4 vCPU and 8 GiB, 60 GB of storage, Ultra Usage, assistant email, and a subdomain. The higher tier is for users whose assistant needs the largest listed hosted resources; the plan descriptions do not give usage quantities for comparing included allowances.

Platforms

Vellum supports Android, iOS, macOS, Windows, web, and an extension, alongside self-hosting. Linux has no shipped desktop client, but Linux users can use the web app, Chrome extension, or self-host the runtime. Support contact is [email protected].

Who it's for

Vellum is aimed at individuals who want a persistent assistant that can work across personal tools and who value approval prompts before sensitive actions. Its cloud hosting and skill customization are most compelling when a user wants an assistant to remain available and adapt to particular tasks. Vellum characterizes itself as a personal assistant, not a shared team or company tool.

Pros and cons

• Pro: Saved workspace context and custom skills support continuity and personalization across conversations.
• Pro: Approval prompts and configurable trust rules give users a way to govern sensitive actions, while Vellum says OAuth tokens are kept from the language model.
• Pro: Cloud plans include isolated environments and private storage, with higher tiers offering larger computers and more storage.
• Con: The Free plan relies on pay-as-you-go credits after initial tasks, so it is not a clearly defined ongoing usage allowance.
• Con: Paid plans begin at $30 a month, and higher resource tiers rise to $100 a month and $200 a month.
• Con: Vellum requires an internet connection to reach a model provider and has no shipped Linux desktop client.

Alternatives

For AI evaluation and observability rather than a personal task assistant, these tools may be a better match:

• Weights & Biases is a freemium option with a free plan and a free trial; its listed platforms include API, Linux, macOS, self-hosted, web, and Windows.
• HELM is a free, open-source Python framework with no paid plans listed, for readers seeking an evaluation framework rather than Vellum's personal assistant.
• Opik offers an open-source core for observability and evaluation that users can download and run locally, making it a fit for readers prioritizing locally run evaluation tooling.
• Braintrust has a free Starter plan with 1 GB processed data, 10,000 scores, 14-day retention, and unlimited users, projects, and datasets, for readers who need those defined evaluation allowances.
• Confident AI has a free plan capped at 2 user seats, 1 project, 5 test runs per week, and 1 GB-month of trace spans; it fits readers whose evaluation work can stay within those limits.
• DeepEval is a free, Apache 2.0-licensed open-source LLM evaluation framework with a local and CI/CD test runner, for readers focused on evaluation workflows.
• Evidently AI offers a fully open-source framework under the Apache 2.0 license, for readers who want open-source AI evaluation tooling.
• Galileo is a freemium alternative with a listed Pro plan at $100 a month, billed yearly.

Readers comparing adjacent categories can also browse AI LLM Evaluation Tools, AI Prompt Generators, AI Prompt Management Software, Prompt Management Software, AI Workflow Automation Software, and AI Agent Platforms.

Verdict

Vellum is a strong choice for individuals who want a configurable assistant with saved context, custom skills, and approval controls across connected services. Its clearest reason to choose it is the combination of task reach and user-controlled permissions; its clearest reasons to look elsewhere are the internet requirement, the absence of a Linux desktop client, and paid plans starting at $30 a month.