Repolode
Submit
vellum-ai

vellum-ai/vellum-assistant

A personal AI assistant that runs on macOS, Windows, iOS, Telegram, Slack and more, with persistent memory and proactive follow-ups.

Open-source alternative to:ChatGPT

View on GitHubWebsite
Stars
1,445
Forks
216
Open issues
184
Last commit
yesterday

What is it?

Vellum Assistant is a personal AI assistant meant to run continuously and act on your behalf across messaging, files, browsing and connected apps. It keeps a memory split into eight types, including episodic, semantic, procedural and prospective memory, and extracts structured items such as preferences, projects and events from conversations. Behavior is defined in editable personality files that the assistant builds during onboarding.

The assistant checks its notes on an hourly cycle, surfaces unfinished or upcoming items, and sends notifications to the appropriate channel. Every tool call runs in a sandbox, actors are classified as guardian, trusted or unknown with restricted access for unknown actors, and credentials are kept in a separate process away from the model. Users can grant computer-use permissions for file, command and browser actions with a choice of one-time, ten-minute or permanent approval.

Model support includes Anthropic, OpenAI, Google Gemini, OpenRouter and any OpenAI-compatible endpoint, with local models through Ollama. It can be used as a managed cloud runtime or self-hosted, with a CLI for installing and managing assistants.

Key features

  • 01Eight memory types with hybrid retrieval and per-user and per-channel isolation
  • 02Proactive hourly review that reminds you of unfinished or due items
  • 03Sandboxed tool execution with permission-gated computer use
  • 04Skills defined by SKILL.md and TOOLS.json that add tools at runtime
  • 05One assistant and one memory across macOS, Windows, iOS, Telegram and Slack
  • 06OAuth integrations with Slack, Notion, Google, HubSpot, Linear and others

Use cases

  • Keeping a personal assistant that remembers your preferences and ongoing projects across devices
  • Receiving reminders and follow-ups for upcoming deadlines through Telegram or Slack
  • Letting an AI agent read files, run commands or drive a browser on your machine with approval
  • Self-hosting a personal AI assistant that uses your own model API keys

Topics

#personal-agent#ai-assistant#agent#memory#claude#openai#telegram-bot#slack#agentic-ai#anthropic

FAQ

What is vellum-assistant?
A personal AI assistant that runs on macOS, Windows, iOS, Telegram, Slack and more, with persistent memory and proactive follow-ups. Vellum Assistant is a personal AI assistant meant to run continuously and act on your behalf across messaging, files, browsing and connected apps. It keeps a memory split into eight types, including episodic, semantic, procedural and prospective memory, and extracts structured items such as preferences, projects and events from conversations. Behavior is defined in editable personality files that the assistant builds during onboarding.
Is vellum-assistant free to use? What license does it have?
vellum-assistant is open source under the MIT license. Check the license terms before using it in commercial products.
Is vellum-assistant still maintained?
Actively maintained. The last commit to vellum-assistant was on Oct 21, 2026.
What can I use vellum-assistant for?
Keeping a personal assistant that remembers your preferences and ongoing projects across devices Receiving reminders and follow-ups for upcoming deadlines through Telegram or Slack Letting an AI agent read files, run commands or drive a browser on your machine with approval Self-hosting a personal AI assistant that uses your own model API keys
Where can I get vellum-assistant?
The source code and installation instructions are on GitHub: https://github.com/vellum-ai/vellum-assistant

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects