Enterprise-grade agentic workflow platform for building and running AI agents, with model management, MCP tool integration and RPA automation.
- Stars
- 665
- Forks
- 92
- Open issues
- 41
- Last commit
- 5 hours ago
What is it?
FLUJO is a local-first AI workspace for building reusable agents. You save model connections (OpenAI, Anthropic, Google Gemini, Azure OpenAI, OpenRouter, or local models through Ollama) and app connections over the Model Context Protocol, then assemble agents either as plain-language steps or as graph flows with AI, tool, static, subflow and finish nodes.
Agents can branch, loop and hand off work between specialists, and a built-in visual debugger lets you set breakpoints and inspect state during execution. The same agents can be used in the Talk chat view, run on schedules, webhooks, file changes or tool results, or be called from external clients through an OpenAI-compatible endpoint and MCP proxy.
Workspace data such as models, agents, chats and settings is stored on the local machine. Cloud providers and connected services receive the requests needed for their operation. The default setup is intended for a single trusted operator on localhost, and network exposure needs separate controls.
Key features
- 01Save model connections for OpenAI, Anthropic, Gemini, Azure OpenAI, OpenRouter and Ollama, with fallback policies.
- 02Connect tools and resources through MCP servers from a registry, a URL, a GitHub repo or a local command.
- 03Build agents as easy step lists or expert visual graphs with branching, loops and subflows.
- 04Debug runs with breakpoints, step-through execution and state inspection.
- 05Trigger agents on schedules, webhooks, file changes, tool results or other flow completions.
- 06Expose agents through an OpenAI-compatible API and an MCP endpoint for external clients.
Use cases
- Prototyping a multi-step agent that combines several models and local tools.
- Running a recurring report or monitoring task that reads files or tool outputs on a schedule.
- Using a configured set of MCP tools from Cline, Cursor or another compatible client.
- Keeping model keys and agent data on a personal machine instead of a hosted platform.
Quick start
- 1
Run the Windows installer or the official install script from the README
- 2
Or, if Node.js 22.23.4+ or 24.22.1+ is installed, run the published package
npx flujo-ai@latest - 3
Open localhost:4200 and keep the FLUJO server running
Things to know
- Some connected tools need extra runtimes or credentials.
- Automations only run while the FLUJO server is running.
- Some features such as Meetings, Personas and Waves are experimental.
Topics
FAQ
- What is FLUJO?
- Open-source local-first workspace for building AI agents visually, connecting models and MCP tools, and running agents in chat or automations. FLUJO is a local-first AI workspace for building reusable agents. You save model connections (OpenAI, Anthropic, Google Gemini, Azure OpenAI, OpenRouter, or local models through Ollama) and app connections over the Model Context Protocol, then assemble agents either as plain-language steps or as graph flows with AI, tool, static, subflow and finish nodes.
- Is FLUJO free to use? What license does it have?
- FLUJO is open source under the MIT license. Check the license terms before using it in commercial products.
- Is FLUJO still maintained?
- Actively maintained. The last commit to FLUJO was on Oct 22, 2026.
- What can I use FLUJO for?
- Prototyping a multi-step agent that combines several models and local tools. Running a recurring report or monitoring task that reads files or tool outputs on a schedule. Using a configured set of MCP tools from Cline, Cursor or another compatible client. Keeping model keys and agent data on a personal machine instead of a hosted platform.
- How do I install FLUJO?
- Run the Windows installer or the official install script from the README Or, if Node.js 22.23.4+ or 24.22.1+ is installed, run the published package: npx flujo-ai@latest Open localhost:4200 and keep the FLUJO server running
- Where can I get FLUJO?
- The source code and installation instructions are on GitHub: https://github.com/mario-andreschak/FLUJO
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
An agent harness for Claude Code and Codex that adds multi-agent swarms, self-learning memory, vector retrieval and cross-machine agent federation.
A multi-agent system built on Context Engineering, with a visual editor for designing, testing and deploying agent workflows.
A zero-code platform that generates production-style AI agents from natural-language descriptions, with built-in tools, memory and orchestration.
A self-hostable open-source visual platform for building automations, integrations, scheduled jobs and HTTP-backed workflows, with AI-assisted flow generation.
Open-source agent runtime and terminal TUI for long-horizon research and file-based work, with ReAct and parallel Agent Team modes.