An open-source, Python-native, node-based workflow engine with a modern React Flow canvas. Build, execute, and manage complex application and AI pipelines visually — with zero enterprise feature gating.
A workflow engine that puts Python first and locks nothing behind a paywall.
Flowy combines the runtime robustness of Node-RED, Meshroom's step-through execution, Langflow's extensible node catalogue, and Zetaforge's container execution — unified in a single Python-backed platform.
Python-Native Backend
Every node's logic is Python. No JavaScript runtime, no YAML-only definitions. The entire Python ecosystem — NumPy, pandas, scikit-learn, transformers — is available natively inside every node.
Visual Step-Through
Play, pause, step, and resume flows node-by-node with real-time status indicators. Yellow for running, green for done, red for failed. Reset and retry individual nodes without restarting the entire flow.
DAG and Non-DAG Flows
Full support for cyclic graphs and iterative loops alongside standard DAGs. Three loop modes — for-each, while-condition, and converge — for building agent patterns, retry logic, and iterative refinement workflows.
No Enterprise Gating
RBAC, audit logs, sharing, SSO — all features available to all users from day one. No "contact sales" for collaboration features. Open source means open source.
Declare nodes in Python. Discover them everywhere.
Flowy's node system is designed around Python conventions. Declare nodes with a decorator or a class, publish them as pip packages, or drop them in a local directory. The catalogue, marketplace, and AI builder handle the rest.
Dual Node Declaration
Two ways to create nodes: class inheritance for complex stateful nodes, or the @flowy.node() decorator for simple functions with automatic type-hint inference. Both produce identical registry entries.
Three-Layer Discovery
Built-in nodes ship with the core package. Community nodes install via pip with entry-point auto-registration. Local nodes live in ~/.flowy/nodes/ for personal utilities and rapid prototyping.
30+ Built-In Nodes
Math, Network, Data, Code, Utility, AI, Flow Control, Debug, and System nodes across 9 categories. Each node ships with typed inputs and outputs, inline documentation, and configurable parameters.
Node Marketplace
A browsable, searchable marketplace UI for discovering, installing, and managing node packages. Filter by category, sort by popularity, and install with one click. Package authors publish via standard pip tooling.
Everything you need to build, run, and share workflows.
A modern canvas, custom dashboards, AI integration, container isolation, collaboration tools, and environmental awareness — all included, all open source.
React Flow Canvas
A modern, performant node graph editor built on React Flow v12 with zoom, pan, minimap, keyboard shortcuts, and multi-input/output port rendering. Handles hundreds of nodes without frame drops.
Custom Dashboards
Nodes produce UI widgets — charts, tables, 3D viewers, maps, code editors — arranged on drag-and-drop dashboards. 11 widget types with automatic data binding from node outputs.
Claude AI Nodes
Built-in nodes for Anthropic Claude API including messages, tool use, and streaming. Plus an AI-powered node builder that generates custom node code from natural language descriptions.
Docker Container Nodes
Execute node processes inside Docker containers for isolation and reproducibility. Pin dependencies, mount volumes, and run untrusted code safely. Container images cached for fast re-execution.
Flow Sharing & Collaboration
Share dialog with granular permissions, editor locking for concurrent access, flow forking with full provenance tracking, inline comments, activity feed, and real-time notifications.
Carbon Impact Tracking
Optional per-node environmental impact metadata including wattage estimates, CO2 calculations, and energy source tags. Aggregate flow-level carbon reporting for sustainability-aware workflows.
Built for people who build things.
Flowy is designed for developers, data scientists, ML engineers, and platform teams who want visual workflow tooling backed by real Python code — not a low-code compromise.
Developers & Tinkerers
Building automation, data pipelines, or AI workflows with a visual tool backed by real code. Every node is Python you can read, debug, and extend.
AI/ML Practitioners
Assembling LLM chains, multi-model pipelines, and agentic workflows with step-by-step visibility. Debug AI workflows node-by-node instead of staring at logs.
Engineering Teams
Shared workflow authoring with RBAC, audit logging, and collaboration features — without paying for enterprise tiers. The full platform, for the full team.
DevOps / Platform Engineers
Wiring together services, APIs, and container-based tasks with a visual orchestrator. Docker container nodes and step-through execution make debugging production pipelines straightforward.
Frequently Asked Questions
Node-RED is JavaScript-based and designed primarily for IoT event wiring. Flowy is Python-native — every node runs Python, giving you access to the entire scientific and AI ecosystem. Flowy also provides a modern React Flow canvas (vs Node-RED's older D3 editor), built-in AI integration with Claude nodes, visual step-through execution for debugging, and DAG + non-DAG flow support with three loop modes for iterative agent patterns.
Many workflow tools lock RBAC, SSO, audit logs, and collaboration features behind expensive enterprise tiers. In Flowy, every feature is available to every user — whether you are a solo developer or a 200-person engineering team. Role-based access control, single sign-on, sharing, audit trails, and flow forking are all included in the open-source release. No "contact sales" buttons.
Yes, two ways. The @flowy.node() decorator lets you turn any Python function into a node with automatic type-hint inference for inputs and outputs. For more complex nodes, extend the base FlowyNode class with full lifecycle control. You can also use the AI-powered node builder to generate node code from a natural language description. Custom nodes can be shared as pip packages or placed in ~/.flowy/nodes/.
Yes. Flowy ships with built-in Claude AI nodes for Anthropic's API — including messages, tool use, and streaming. Non-DAG flow support with while-condition loops enables agentic patterns where an LLM decides whether to continue iterating. Step-through execution lets you inspect every intermediate result in an AI pipeline, making it far easier to debug multi-step LLM workflows than reading logs.
Like a debugger for workflows. Click Play to run the entire flow, or Step to advance one node at a time. Each node shows real-time status: yellow while running, green when complete, red on failure. You can pause mid-flow, inspect intermediate data on any edge, reset individual nodes to retry them, and resume from any point. This makes complex pipelines transparent and debuggable without adding print statements everywhere.
Local-first. Flowy runs on any machine with Python 3.10+ and Node.js. Your flows, data, and credentials stay on your hardware. An AWS Amplify deployment option is planned for teams that want a hosted instance, but the core product is designed to run locally with no cloud dependency. You own your data and your workflows.
Build, execute, and manage workflows visually.
Flowy is coming soon. Register your interest to get notified when the open-source release is available.
AI software engineering and innovation
for ambitious organisations