OpenVault AI

unlisted

by Or Shemtov

AI assistant for note-aware chat, editable conversations, custom agents, and flexible model providers.

2 starsUpdated 29d agoMIT
View on GitHub

OpenVault AI

OpenVault AI is an Obsidian assistant for talking to your notes, organizing context, and building repeatable workflows with markdown-defined agents, skills, and commands.

OpenVault AI logo

Release CI License GitHub stars

Issues PRs welcome Obsidian plugin

OpenVault AI screenshot

Features

  • Chat with a single note, a folder, selected notes, or the whole vault
  • Reference notes and folders directly with @mentions
  • Run slash commands from markdown-defined command files
  • Switch between built-in agents like ask and edit
  • Add your own agents in AI/Agents/<agent-name>/AGENT.md
  • Add reusable skills in AI/Skills/<skill-name>/SKILL.md
  • Persist conversations in the vault as markdown files
  • Store long-term memory entries in the vault for preferences, facts, and lessons
  • Switch between Ollama, OpenRouter, OpenAI, and Anthropic from the plugin UI
  • Review tool usage and assistant context directly in the UI

How It Works

OpenVault AI keeps the assistant close to your vault instead of hiding behavior behind a remote service.

  • Notes stay addressable as files and folders
  • Agents, skills, and commands are editable markdown files in the vault
  • Conversations are saved as markdown notes
  • Long-term memory is stored as markdown entries in the vault
  • Provider settings live in the plugin settings for the current vault

Memory

OpenVault AI supports two kinds of memory:

  • Conversation memory: active chats and previous conversations stored in the vault
  • Long-term memory: saved preferences, facts, and lessons that the assistant can reuse later

This makes it possible to keep context across sessions while still keeping the data visible and editable.

Agents, Skills, And Commands

The plugin supports built-in behavior plus vault-defined extensions.

Agents

Built-in agents:

  • ask
  • edit

Vault-defined content is loaded from the AI/ prefix by default.

Custom agents live in:

AI/Agents/<agent-name>/AGENT.md

Skills

Custom skills live in:

AI/Skills/<skill-name>/SKILL.md

Commands

Custom slash commands live in:

AI/Commands/<command-name>.md

Installation

Community plugins

Once approved in the Obsidian marketplace:

  1. Open Obsidian
  2. Go to Settings -> Community plugins
  3. Search for OpenVault AI
  4. Install the plugin
  5. Enable it

Manual installation

  1. Download main.js, manifest.json, and styles.css from a GitHub release
  2. Create this folder inside your vault:
<vault>/.obsidian/plugins/openvault-ai/
  1. Copy the release files into that folder
  2. Reload Obsidian and enable the plugin

Usage

Open the assistant

Use one of these commands from the command palette:

  • Open assistant
  • Toggle assistant

Configure a provider

The settings tab lets you configure:

  • Ollama base URL
  • OpenRouter base URL and API key
  • OpenAI base URL and API key
  • Anthropic base URL and API key

Ask about notes

Inside the assistant you can reference vault content directly in the prompt.

Examples:

  • @Daily/2026-04-24.md summarize this note
  • @Projects/Roadmap/ what are the open decisions here?
  • @all compare the main themes across my recent planning notes

Privacy And Data Handling

  • Ollama requests stay on your local machine unless your Ollama server is remote
  • OpenRouter, OpenAI, and Anthropic send request data to external services you configure
  • Prompts may include note content that you explicitly reference or that the plugin retrieves for the active request
  • API keys are stored in the plugin's local Obsidian data file for the current vault
  • This plugin does not include telemetry or analytics collection

You are responsible for choosing which provider to use for a given vault and what content you send to external APIs.

Contributing

Development setup, local workflows, and contributor checks live in CONTRIBUTION.md.

License

MIT

For plugin developers

Search results and similarity scores are powered by semantic analysis of your plugin's README. If your plugin isn't appearing for searches you'd expect, try updating your README to clearly describe your plugin's purpose, features, and use cases.