ThoughtForge

A local-first AI planning assistant. Extracts action items from meeting transcripts, manages tasks on a kanban board, and uses a local LLM (via LM Studio) for smart planning — all without sending data to the cloud.

Screenshots

View	Dark	Light
Dashboard
Kanban Board
Space & Notes
Weekly Review
AI Chat
Settings

Features

• Dashboard — smart filters: My Tasks, Overdue, Due Today, Blocked, by project
• Kanban Board — Time-based, Status, Day-by-day, and Calendar views
• Spaces — per-project workspace with notes, tasks, booked hours, and semantic search
• AI Chat — daily/weekly planning via local LLM; changes confirmed before writing
• Document Processing — upload transcripts (.txt, .md, .pdf), auto-extract action items
• Time Tracking — book hours per space, weekly timesheet grid, daily activity chart
• Semantic Search — per-space vector index powered by LM Studio embeddings
• Model Recommendations — RAM-aware model suggestions with LM Studio copy-paste commands
• 100% Local — no network calls except localhost. No telemetry.

Architecture

Layer	Technology
App framework	Tauri 2.x (Rust backend)
Frontend	React 19 + TypeScript
Styling	Tailwind CSS
State	Zustand
LLM	OpenAI-compatible API via LM Studio
Storage	Markdown + JSON files

Storage layout

~/Documents/ThoughtForge/
  spaces/
    {id}/
      space.json        # metadata (name, color, time entries)
      notes/
        {note-id}.md    # YAML frontmatter + markdown body
      tasks/
        {task-id}.md    # YAML frontmatter + notes body
      search_index.json # per-space vector index
  boards/
    kanban.md
  uploads/
    transcripts/
    documents/
  config.yaml

Prerequisites

• macOS / Windows / Linux
• Rust (1.70+): curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
• Node.js (20+): brew install node
• LM Studio: lmstudio.ai — for AI features

Setup

git clone https://github.com/larsroettig/thoughtforge.git
cd thoughtforge
npm install

Development

npm run tauri dev

Production Build

npm run tauri build

Produces a .app bundle and .dmg installer in src-tauri/target/release/bundle/.

LM Studio Setup

1. Install LM Studio and load models:

   lms load qwen2.5-14b-instruct        # chat (recommended for 16–32 GB RAM)
   lms load nomic-embed-text-v1.5       # embeddings (for semantic search)
   

2. In ThoughtForge → Settings → verify URL is http://localhost:1234
3. Select your chat model from the dropdown

See Settings → Model Recommendations for RAM-matched model suggestions.

AI Chat

The chat connects to your local LLM and can propose task modifications:

1. Ask "plan my day" or "change priority of X to high"
2. The AI responds with proposed changes
3. A confirmation card appears — review each change, apply or skip individually
4. The AI never writes to your vault without explicit confirmation

Slash Commands

Command	Description
/plan-day	Daily plan with priorities and time blocks
/plan-week	Day-by-day weekly breakdown
/status	Overview of all projects and blockers
/blocked	Show blocked or overdue items
/extract	Extract action items from pasted text
/summarize	Summarize a project
/prioritize	Suggest priority ordering

MCP Server (Claude Desktop Integration)

ThoughtForge ships a built-in Model Context Protocol server. Enable it in Settings → MCP & Integrations, then add the server to claude_desktop_config.json:

{
  "mcpServers": {
    "thoughtforge": {
      "command": "/Applications/ThoughtForge.app/Contents/MacOS/vaultmind-mcp",
      "args": ["--stdio"]
    }
  }
}

Available tools: list_tasks, create_task, update_task, delete_task, list_spaces, list_notes, create_note, search_notes, list_goals, create_goal, update_goal, delete_goal.

See docs/mcp.md for the full tool reference and HTTP transport setup.

Documentation

• User Guide — getting started, all features explained
• MCP Server — Claude Desktop integration, tool reference, security
• GitHub Pages — landing page

Security & Permissions

ThoughtForge requests only the Tauri capabilities it needs. See PERMISSIONS.md for a full table of every granted permission, the code that uses it, and the rationale.

Key points:

• No network calls outside localhost — SSRF guard in src-tauri/src/llm.rs blocks all non-localhost hosts
• No telemetry — the app cannot phone home
• Frontend never touches the filesystem directly — all I/O goes through validated Rust commands
• MCP bearer token compared with constant-time equality

Testing

npm test                          # Vitest unit tests
npm run test:e2e                  # Playwright E2E (requires: npm run dev in another terminal)
cd src-tauri && cargo test        # Rust unit tests

Author

Built by Lars Roettig.

License

MIT