AI Copilot

Your intelligent writing and thinking assistant for Obsidian.

AI Copilot brings Notion AI-like capabilities directly into your vault. Chat with your notes, transform selected text, search the web, manage personas, and extend functionality with MCP servers and custom skills — all from a native-feeling sidebar.

✨ Features

🗨️ AI Chat Sidebar

• Context-aware conversations — Chat with your current note, selected text, or entire vault.
• @ mentions — Reference specific files inline to pull them into chat context.
• Right-click integration — Send files, folders, or selections to chat from context menus.
• Multi-session support — Pin and switch between multiple chat sessions with tabs.
• Project scoping — Organize chats by project with scoped folders, tags, and system prompts.
• Auto-focus input — The chat input is ready for typing whenever you open the panel.

🤖 Multiple AI Providers

Connect to your preferred LLM provider:

• OpenAI (GPT-4o, GPT-4.1, o3-mini, etc.)
• Anthropic (Claude 3.5/4 Sonnet, Opus, etc.)
• Google Gemini (Gemini 2.5 Pro/Flash, etc.)
• Groq (Llama, Mixtral, etc.)
• Ollama (Local models — fully offline, privacy-first)

✍️ Text Transform Actions

Select text and run AI-powered transformations from the command palette or right-click menu:

🛠️ Built-in Tools (Function Calling)

The AI can autonomously use tools during conversation:

• create_note / append_to_note / read_note / edit_note — Full vault CRUD operations
• list_folder — Browse vault structure
• web_search — Search the web and bring results into chat
• get_youtube_transcript — Fetch and summarize YouTube video transcripts
• read_pdf — Extract text from PDF files in your vault
• summarize_url — Fetch and summarize any web page
• search_vault_by_date — Find notes by creation or modification date
• save_memory / list_memories — Long-term memory across sessions
• save_summary_as_note — Save AI-generated summaries directly to your vault

🔍 Vault QA & Embeddings

• Semantic search over your entire vault using vector embeddings.
• Auto-indexing — Automatically re-index notes on startup or change.
• Supports OpenAI and Ollama embedding providers.
• Configurable index exclusions for folders you don't want indexed.

🎭 Personas

• Create multiple AI personalities with custom system prompts.
• Persona soul & memory — Each persona remembers past interactions and learnings.
• Persona-specific mistake tracking — AI learns from its errors per persona.
• Switch personas on-the-fly from the chat interface.

📂 Projects

• Define scoped contexts with specific folders, tags, and system prompts.
• Override the model per project for cost/performance optimization.
• Create projects from settings or directly inline from the chat interface.
• Automatically filter relevant notes based on project scope.

🔌 MCP Server Integration

Connect to local Model Context Protocol (MCP) servers to extend AI capabilities:

• Auto-detect server directories and .env files.
• Resolve bare command names across system PATH.
• Shell wrapper support for complex server startup.
• Configure command, arguments, environment variables, and working directory per server.

🧠 Skills System

• Point to a folder of SKILL.md files to extend AI behavior.
• Skills are automatically discovered, indexed, and injected based on query relevance.
• Enable/disable individual skills; mark skills as mandatory for always-on injection.

⚡ Custom Actions

• Create your own text-transformation commands with custom prompt templates.
• Use {{selection}} placeholder to reference highlighted text.
• Custom actions appear in the Obsidian command palette for quick access.

🖱️ Right-Click Context Menu

Full AI Copilot submenu in the editor right-click menu:

• Fix grammar, reformat, summarize, simplify, shorten, lengthen, change tone, ELI5.
• Send files and folders to chat from the file explorer.

⌨️ Keyboard Shortcuts

🛠️ Installation

From Obsidian Community Plugins

1. Open Settings → Community plugins → Browse.
2. Search for AI Copilot.
3. Click Install, then Enable.

Manual Installation

1. Download the latest release from GitHub Releases.
2. Copy main.js and manifest.json to your vault's plugin folder:

   .obsidian/plugins/ai-copilot/
   

3. Reload Obsidian and enable the plugin.

Build from Source

git clone https://github.com/alexjiaguo/obsidian-ai-copilot.git
cd obsidian-ai-copilot
npm install
npm run build

⚙️ Configuration

1. Open Settings → AI Copilot.
2. Select your AI provider and enter your API key.
3. (Optional) Customize:
   • Personas — Create AI personalities with custom system prompts.
   • Projects — Scope conversations to specific folders and tags.
   • Custom actions — Build your own text transformation commands.
   • MCP servers — Connect to local MCP servers for extended tooling.
   • Skills — Point to a skills folder for automatic prompt augmentation.
   • Vault QA — Configure embedding provider and auto-indexing.

🏗️ Architecture

main.ts                          → Plugin entry point, commands, menus
src/
├── views/
│   ├── AIChatView.ts            → Obsidian ItemView wrapper
│   ├── ChatApp.svelte           → Tab management, session lifecycle
│   └── ChatView.svelte          → Chat UI, message processing
├── components/
│   ├── ChatInput.svelte         → Auto-resizing input with @ mentions
│   ├── MessageBubble.svelte     → Message rendering with markdown
│   └── ProjectSelector.svelte   → Inline project creation/selection
├── services/
│   ├── APIService.ts            → Multi-provider LLM abstraction
│   ├── ToolManager.ts           → Function calling (17 built-in tools)
│   ├── MCPClientService.ts      → MCP server connections
│   ├── SkillService.ts          → Skill discovery and injection
│   ├── VaultQA.ts               → Embedding-based semantic search
│   ├── MemoryService.ts         → Persistent cross-session memory
│   ├── PersonaSoulService.ts    → Persona-specific memory and soul
│   ├── ContextManager.ts        → File/folder search and context
│   ├── EditorHandler.ts         → Active editor tracking
│   ├── RelevantNotes.ts         → Context-aware note suggestions
│   ├── WebSearch.ts             → DuckDuckGo web search
│   ├── YouTubeTranscriber.ts    → YouTube transcript extraction
│   ├── PDFService.ts            → PDF text extraction
│   ├── ContentExtractor.ts      → URL content extraction
│   └── EmbeddingService.ts      → OpenAI/Ollama embedding client
└── settings/
    ├── Settings.ts              → Type definitions and defaults
    └── SettingsView.svelte      → Full settings UI

🔒 Privacy

• Your data stays local unless you choose to send it to an external API.
• Use Ollama for fully offline, on-device AI with zero data leaving your machine.
• API keys are stored locally in your Obsidian vault settings.
• No telemetry, no tracking, no analytics.

🤝 Contributing

Contributions are welcome! Feel free to open issues or submit pull requests.

📄 License

MIT