DISCOVER THE FUTURE OF AI AGENTS

NodeTool

Added May 4, 2026
Agent & Tooling
Open Source
PythonTypeScriptNode.jsWorkflow AutomationElectronLarge Language ModelsMultimodalRAGAI AgentsWeb ApplicationCLIAgent & ToolingModel & Inference FrameworkAutomation, Workflow & RPAKnowledge Management, Retrieval & RAGComputer Vision & Multimodal

A node-based visual AI workflow and LLM Agent builder with local model support and multimodal orchestration across desktop, web, CLI, and mobile.

NodeTool is a local-first, node-based visual platform for building AI workflows and LLM Agents. It features a drag-and-drop node editor with 100+ built-in node types and tools, supporting unified processing of text, image, video, and audio.

Core Capabilities

  • Visual Workflow Engine: Drag-and-drop node editor with type-safe connections, async execution engine with per-node real-time preview, exportable as interactive Mini-Apps
  • AI Agent System: Built-in Agent framework with task planning and tool calling, 100+ built-in tools, interactive Agent chat mode in CLI
  • Model Integration: Local models (Ollama, MLX, GGUF/GGML) for fully offline operation; cloud providers (OpenAI, Anthropic, Gemini, Replicate, FAL, ElevenLabs, HuggingFace) via BYOK
  • RAG & Knowledge Management: Built-in document indexing and semantic search using SQLite-vec vector database with zero external dependencies
  • Multimodal Processing: Text generation, image generation/transformation (FLUX, NanoBanana), video processing, speech transcription and synthesis
  • Extensions & Sandboxing: TypeScript custom nodes (node-sdk BaseNode), Python custom nodes (PythonStdioBridge lazy-loading worker), Docker-isolated code execution sandbox

Product Forms

  • NodeTool Studio: Local desktop app with local model support, data never leaves the machine
  • NodeTool Cloud: Hosted version, browser-accessible, no installation or GPU required
  • Both share the same open-source codebase with seamless workflow portability

Architecture TypeScript monorepo managed by Turborepo (28 backend packages). Key packages: kernel (DAG orchestration), node-sdk (BaseNode class & node registry), agents (task planning & tools), runtime (processing context & LLM providers), websocket (HTTP + WebSocket dual-protocol server), vectorstore (SQLite-vec), code-runners (Docker sandbox). Frontend: React + Vite + MUI + React Flow. Desktop: Electron 39 (embedded Node 24). Mobile: React Native + Expo. Runtime: Node.js 24+ (native async), Python 3.11 optional. Testing: Vitest + Playwright.

Installation

  • Desktop: Download from official site (Windows: NVIDIA GPU 4GB+ VRAM; macOS: M1+ 16GB+ RAM; Linux: NVIDIA GPU or Flatpak)
  • CLI: npm install -g @nodetool-ai/cli — supports nodetool serve, nodetool-chat --agent, nodetool workflows run <file.ts>
  • Source: nvm use && npm install && npm run build:packages && npm run dev
  • Server deployment: Docker, RunPod, Google Cloud Run, or self-hosted

Use Cases: Privacy-compliant scenarios (healthcare, finance), multi-step AI Agent automation, document intelligence & RAG, media generation pipelines, data processing & automated reporting, workflow Mini-Apps sharing.

Current Version: v0.7.0-rc.23 (RC stage), AGPL-3.0 license. Primary language: TypeScript (96%).

Related Projects

View All

STAY UPDATED

Get the latest AI tools and trends delivered straight to your inbox. No spam, just intelligence.