Mentat

approved

by Mentat Team

Your local agentic cognitive partner for Obsidian, managing structured knowledge, local command operations, and deep mathematical reasoning. - This plugin has not been manually reviewed by Obsidian staff.

โ†“ 115 downloadsUpdated 29d agoMIT

Mentat - Obsidian Plugin

Your local agentic cognitive partner for Obsidian. Mentat is an AI agent that lives in your vault โ€” it reads, writes, searches, researches, and summarizes your notes through conversation.

What you can do with it

  • ๐Ÿ’ฌ Chat with your vault: "Find my notes about RAG from last week and summarize them"
  • ๐Ÿ” Deep research: Create structured research plans, search the web, fetch pages, and save findings
  • โœ๏ธ Write and edit notes: "Write a technical summary of this paper to Research/"
  • ๐Ÿ› ๏ธ Automate tasks: Bulk operations, frontmatter updates, file moves with link fixing
  • ๐ŸŽจ Two UI themes: Classic bubble chat or terminal-style timeline โ€” switch anytime
  • ๐Ÿงฉ Extensible: Extension system, skill framework, MCP support

AI Providers

ProviderChat + ToolsEmbeddingLocal
OpenAI (and compatible)โœ…โœ…โŒ
Anthropic Claudeโœ…โŒโŒ
OllamaโŒ (no tool support)โœ…โœ…

Set up in Settings โ†’ Mentat โ†’ Add AI Provider.

Quick start

# Install
cd /path/to/vault/.obsidian/plugins
git clone https://github.com/Secbone/obsidian-mentat mentat
cd mentat && npm install && npm run build

# Enable in Obsidian: Settings โ†’ Community plugins โ†’ Mentat
# Configure: Settings โ†’ Mentat โ†’ Add AI Provider
# Open chat: Ctrl/Cmd+P โ†’ "Open AI Chat"

Development

npm run build        # Type check + bundle โ†’ dist/
npm run deploy:dev   # Copy to ~/Documents/obsidian/.obsidian/plugins/
npm run lint         # ESLint
npm test             # Vitest

Project structure

src/
โ”œโ”€โ”€ agents/          # Agent runtime (BaseAgent, Compactor, events)
โ”œโ”€โ”€ chat/            # Chat orchestration (session, query, subagents)
โ”œโ”€โ”€ context/         # Message windowing, token estimation
โ”œโ”€โ”€ extensions/      # Extension system (EventBus, ExtensionManager)
โ”œโ”€โ”€ providers/       # AI providers (OpenAI, Anthropic, Ollama)
โ”œโ”€โ”€ prompts/         # System prompts, skill prompts, vault context
โ”œโ”€โ”€ settings/        # Plugin settings & settings UI
โ”œโ”€โ”€ skills/          # Skill framework (registry, executor, MCP, strategies)
โ”œโ”€โ”€ ui/              # Themed UI (ChatView, bubble theme, terminal theme)
โ””โ”€โ”€ types/           # Shared type definitions

In-depth docs

DocWhat it covers
src/ui/themes/README.mdTheme system architecture, creating new themes
src/ui/README.mdUI system overview, CSS conventions, icon sizing
src/agents/README.mdAgent system, events, streaming, tool execution
src/skills/README.mdSkill framework internals, MCP integration, custom skills
skills/README.mdAdding and documenting skills
prompts/README.mdPrompt system, template variables
docs/theme-system-refactor.mdTheme system refactoring history

Roadmap

  • Phase 1: Agent system โ€” conversation, tool calling, multi-turn reasoning
  • Phase 2: Skill framework โ€” 12 built-in skills, progressive/native/auto modes, MCP
  • Phase 3: Themed UI โ€” BubbleTheme, TerminalTheme, smart scroll, terminal presets
  • Phase 4: Extension system v1 + automatic context compaction
  • Phase 4.5: EventBus architecture (AsyncGenerator โ†’ EventBus, AgentContext.confirmHandler removed)
  • Custom skills: user-defined skills via .obsidian/plugins/mentat/skills/ with implementation.js
  • Phase 5: Tree-structured sessions with branching and comparison
  • Phase 6: Extension system v2 โ€” dynamic loading, permission sandbox
  • Phase 7: i18n, plugin marketplace, community extension registry

Built with

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.