Intelligence Assistant for Obsidian

Turn your vault into an Obsidian-Native Agentic Workspace. Intelligence Assistant is an autonomous AI agent that deeply understands your vault's structure, links, and canvas. It can independently plan and execute complex knowledge tasks—from organizing folders to building mind maps—all while ensuring safety through a Proposal-First execution model.

Desktop only. Requires Obsidian v1.7.2+.

🚀 The Agentic Evolution

This plugin has evolved beyond a simple chat interface into a full SPAR (Sense-Plan-Act-Reflect) autonomous loop:

• 📡 Sense: The Agent perceives your active note, graph neighbors, directory structure, RAG snippets, and its own long-term memories.
• 🧠 Plan: Using advanced reasoning (Chain-of-Thought), the Agent formulates a multi-step strategy and provides a transparent Task Checklist.
• 🛠️ Act: It independently executes tools (Vault, Canvas, MCP, Web, CLI) to gather information or prepare modifications.
• 🔄 Reflect: It evaluates results and refines its approach. Any vault changes are presented as Batch Write Proposals for your one-click approval.

✨ Feature Highlights

• Built-in Expert Team — Get started instantly with pre-configured personas: Librarian (Organization), Architect (Canvas & Mapping), Researcher (Synthesis), and Scribe (Writing & Vision).
• Spatial & Visual Intelligence — Independent read/write access to Obsidian Canvas (.canvas) files. Use Multi-modal Vision to turn sketches or screenshots into notes.
• Cognitive Memory System — Long-term associative memory via a local vector store. The Agent remembers your preferences and key research findings across conversations.
• Modern Logical UI — A redesigned Execution Timeline that merges reasoning and actions into a single flow. Compact tool pills save space while remaining fully inspectable.
• Vault-Native Tools — Specialized tools for Properties (YAML) management, batch file operations (move, create, delete), and directory-tree awareness.
• Multiple LLM Providers — OpenAI, Anthropic (Claude 3.5), Google Gemini, DeepSeek, Ollama (local), OpenRouter, and SAP AI Core.
• MCP & OpenAPI Integration — Seamlessly connect Model Context Protocol servers or any OpenAPI-spec service to expand your Agent's toolset.
• Privacy First — Your notes stay local. Indexing (RAG) uses a local vector store with configurable embedding providers.

🚀 Quick Start

📦 Community Plugin (recommended)

1. Open Settings → Community plugins in Obsidian.
2. Browse for "Intelligence Assistant" and install.
3. Enable the plugin, then open Settings → Intelligence Assistant to configure providers.

📋 Requirements

Requirement	Details
Obsidian	v1.7.2 or later (desktop only)
LLM API key	OpenAI, Anthropic, Google, DeepSeek, etc.

🤖 Agents

Defined under Settings → Agents. Each agent has:

• System prompt, tool permissions, MCP server access
• Model strategy: default, chat-view (follow chat selector), or fixed
• Capabilities: RAG, Web Search, Long-term Memory
• Custom icon and display name

Agent mode uses native function calling (OpenAI) with automatic fallback to text-based tool parsing for other providers.

🧭 Agentic Scenarios

Command Palette (Cmd+P) and context menus expose task-oriented Agent entry points:

• Folder Organizer — proposes folder cleanup, moves, and index/MOC candidates.
• Project Brief — summarizes project status and next actions from a note or folder.
• Weekly Review — scans daily/project context to draft accomplishments and next-week plans.
• Research Brief — turns notes into a sourced research summary with a reading queue.
• Link/Tag Doctor — diagnoses inconsistent tags and orphaned concepts.
• Writing Scribe — creates drafts from source material and understands image sketches.

🔌 Tooling & Integration

📁 Vault Mastery

• Batch Proposals: Confirm multiple file moves or edits in a single click.
• Property Doctor: Automatically standardize tags, dates, and custom fields.

🎨 Canvas Support

• Independent access to .canvas JSON. The Agent can place cards, link files, and organize your visual workspace.

🌐 Beyond the Vault

• Web Search: 8 supported providers for real-time external grounding.
• MCP & OpenAPI: Connect external data sources and automation scripts.

🔐 Capabilities & Data Access

This is an agentic plugin, so it deliberately uses a few capabilities beyond the standard vault API. Each is opt-in and used only for the feature listed. Nothing is collected or transmitted by the plugin itself — network traffic only goes to the LLM/search providers and MCP/OpenAPI endpoints you configure.

Capability	Why it's needed	Scope / safeguards
Shell execution (child_process)	Running CLI tools and MCP stdio servers that you add	Only commands you configure are run; nothing executes without an explicit tool/server definition
Direct filesystem access (fs)	Locating and probing external executables (e.g. npx, python) when launching MCP servers	Read-only access() checks for resolving command paths; all vault content is read/written exclusively through the Obsidian API
Environment variables	Spawned tools/servers need a working environment (PATH, HOME, locale, proxy, language toolchains)	An allowlist is forwarded — unrelated secrets and identity variables are not passed through. Anything extra a tool needs is set explicitly in that tool's own env config
Vault enumeration (getMarkdownFiles)	Building the local RAG index	Runs only when you trigger indexing; respects your include/exclude settings
Clipboard (navigator.clipboard)	"Copy" buttons in chat, plus optional pbcopy/pbpaste CLI presets	Write-on-click only; presets are off unless you add them
JSON-schema validation (ajv)	Validating tool/agent argument schemas (also used internally by the MCP SDK)	Compiles schemas to validators in-memory; processes only your tool definitions

Vault files are read and written only through the official Obsidian API (vault.read/cachedRead/modify/create). The capabilities above never touch vault content directly.

🌐 Internationalization

Supports all 46 Obsidian languages, including full localization for the Command Palette and Context Menus.

中文文档 | English

🛠️ Development

npm install          # install dependencies
npm run dev          # development build + file watcher
npm run lint         # ESLint check
npm run test         # Jest test suite
npm run build        # production bundle

📖 Documentation

Document	Description
README-zh.md	Chinese README (中文说明)
docs/README.md	Documentation index
docs/architecture/overview-en.md	Architecture overview
docs/project/project-guide-en.md	Developer project guide
docs/reference/project-structure.md	Full source tree reference

Contributions, issues, and feature requests are welcome — open a PR or discussion.