Your AI, your rules. Nodetool Agents run on your devices, where your data lives. Transparent. Portable. Under your control.

🎯 Privacy by design • 🔓 Own your stack • 🚀 Production ready

• How It Works
• Principles
• What NodeTool Is
• What NodeTool Is Not
• Who It’s For
• Roadmap
• Quick Start
• Bring Your Own Providers
• Install Node Packs in the App
• Community
• 🛠️ Development Setup
• Run Backend & Web UI
• Testing
• Troubleshooting
• Contributing
• License
• Get in Touch

Get from idea to production in three simple steps:

1. 🏗️ Build — Drag nodes to create your workflow—no coding required.
2. ⚡ Run — Test locally. Your data stays on your machine by default.
3. 🚀 Deploy — Ship with one command to RunPod or your own cloud.

• Local‑first, by design.
• No data leaves without explicit consent.
• Open and portable. Graphs are open and run anywhere.
• One graph. One runtime. Visual and headless share the same semantics.
• Typed tools. Small, composable nodes with explicit I/O.
• Reproducible runs. Seeds, inputs, and artifacts are captured.
• No black boxes. Inspect every step as it runs.
• Fast on your hardware. Optimized for small GPUs.

• Visual graph editor + runnable runtime (desktop + headless)
• Execute via CLI, HTTP API, or WebSocket; graphs in JSON/DSL
• Local models (Ollama/HF) + optional cloud (OpenAI/Gemini/Anthropic/Replicate/Fal/Hugging Face Hub)
• Deploy to laptop/server, Runpod, or Google Cloud

• Managed SaaS with SLAs or multi‑tenant hosting
• Compliance‑grade policy or audit tooling
• An autoscaling orchestration platform
• A one‑click content toy

Indie hackers, game devs, AI enthusiasts, creative professionals, agencies, and studios who want to build and run their own pipelines—with control and local ownership of data and IP.

• Creative pipelines: ready‑to‑use templates for thumbnails, game sprites, ad variants, research briefs.
• Packaging: one‑click bundles for sharing pipelines with collaborators or clients.
• Video tooling: timeline nodes, effects, and batch rendering.
• Community packs: growing registry of open nodes for creative tasks (effects, generators, integrations).

1. Download and install NodeTool
2. Launch the app
3. Download models
4. Start with a template or create from scratch
5. Drag, connect, run—see results instantly

Connect to any AI provider. Your keys, your costs, your choice.

✅ Integrated Providers: OpenAI • Anthropic • Hugging Face • Groq • Together • Replicate • Cohere • + 8 more

Set provider API keys in Settings → Providers.

Install and manage packs directly from the desktop app.

• Open Package Manager: Launch the Electron desktop app, then open the Package Manager from the Tools menu.
• Browse and search packages: Use the top search box to filter by package name, description, or repo id.
• Search nodes across packs: Use the “Search nodes” field to find nodes by title, description, or type. You can install the required pack directly from node results.

Open source on GitHub. Star and contribute.

💬 Join Discord — Share workflows and get help from the community

🌟 Star on GitHub — Help others discover NodeTool

🚀 Contribute — Help shape the future of visual AI development

Follow these steps to set up a local development environment for the entire NodeTool platform, including the UI, backend services, and the core library (nodetool-core). If you are primarily interested in contributing to the core library itself, please also refer to the nodetool-core repository for its specific development setup using Poetry.

• Python 3.11: Required for the backend.
• Conda: Download and install from miniconda.org.
• Node.js (Latest LTS): Required for the frontend. Download and install from nodejs.org.

# Create and activate the Conda environment
conda create -n nodetool python=3.11 -y
conda activate nodetool

# Install essential system dependencies via Conda
conda install -c conda-forge ffmpeg cairo x264 x265 aom libopus libvorbis lame pandoc uv -y

These are the essential packages to run NodeTool.

Make sure to activate the conda environment.

# Install nodetool-core and nodetool-base
# On macOS / Linux / Windows:
uv pip install git+https://github.com/nodetool-ai/nodetool-core
uv pip install git+https://github.com/nodetool-ai/nodetool-base

If you're working in this monorepo and want live-editable installs:

# From the repository root
conda activate nodetool
uv pip install -e ./nodetool-core
uv pip install -e ./nodetool-base

NodeTool's functionality is extended via packs. Install only the ones you need.

NOTE:

• Activate the conda environment first
• Use uv for faster installs.

Prefer the in‑app Package Manager for a guided experience. See Install Node Packs in the App. The commands below are for advanced/CI usage.

# List available packs (optional)
nodetool package list -a

# Example: Install packs for specific integrations
uv pip install git+https://github.com/nodetool-ai/nodetool-fal
uv pip install git+https://github.com/nodetool-ai/nodetool-replicate
uv pip install git+https://github.com/nodetool-ai/nodetool-elevenlabs

Note: Some packs like nodetool-huggingface may require specific PyTorch versions or CUDA drivers. Use --index-url to install:

1. Check your CUDA version:

nvidia-smi

2. Install PyTorch with CUDA support first:

# For CUDA 11.8
uv pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118

# For CUDA 12.1-12.3 (most common)
uv pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121

# For CUDA 12.4+
uv pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu124

3. Install GPU-dependent packs:

# Use --extra-index-url to access both PyPI and PyTorch packages
uv pip install --extra-index-url https://download.pytorch.org/whl/cu121 git+https://github.com/nodetool-ai/nodetool-huggingface

4. Verify GPU support:

python -c "import torch; print(f'CUDA available: {torch.cuda.is_available()}')"

If you see "bitsandbytes compiled without GPU support", reinstall it:

uv pip uninstall bitsandbytes
uv pip install bitsandbytes

If PyTorch shows CPU-only version, make sure you used the correct url from step 2.

Use --extra-index-url (not --index-url) when installing from git repositories to avoid missing dependencies

Ensure the nodetool Conda environment is active.

Option A: Run Backend with Web UI (for Development)

This command starts the backend server:

# On macOS / Linux / Windows:
nodetool serve --reload

Run frontend in web folder:

cd web
npm install
npm start

Access the UI in your browser at http://localhost:3000.

Option B: Run with Electron App

This provides the full desktop application experience.

Configure Conda Path: Ensure your settings.yaml file points to your Conda environment path:

• macOS/Linux: ~/.config/nodetool/settings.yaml
• Windows: %APPDATA%/nodetool/settings.yaml

CONDA_ENV: /path/to/your/conda/envs/nodetool # e.g., /Users/me/miniconda3/envs/nodetool

Build Frontends: You only need to do this once or when frontend code changes.

# Build the main web UI
cd web
npm install
npm run build
cd ..

# Build the apps UI (if needed)
cd apps
npm install
npm run build
cd ..

# Build the Electron UI
cd electron
npm install
npm run build
cd ..

Start Electron:

cd electron
npm start  # launches the desktop app using the previously built UI

The Electron app will launch, automatically starting the backend and frontend.

pytest -q

cd web
npm test
npm run lint
npm run typecheck

cd electron
npm run lint
npm run typecheck

• Node/npm versions: use Node.js LTS (≥18). If switching versions:

  rm -rf node_modules && npm install

• Port in use (3000/8000): stop other processes or choose another port for the web UI.
• CLI not found (nodetool): ensure the Conda env is active and packages are installed; restart your shell.
• GPU/PyTorch issues: follow the CUDA-specific steps above and prefer --extra-index-url for mixed sources.

We welcome community contributions!

1. Fork the repository
2. Create a feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

Please follow our contribution guidelines and code of conduct.

AGPL-3.0 — True ownership, zero compromise.

Tell us what's missing and help shape NodeTool

✉️ Got ideas or just want to say hi?
hello@nodetool.ai

👥 Built by makers, for makers
Matthias Georgi: matti@nodetool.ai
David Bührer: david@nodetool.ai

📖 Documentation: docs.nodetool.ai
🐛 Issues: GitHub Issues

NodeTool — Build agents visually, deploy anywhere. Privacy first. ❤️