ByteDance has open-sourced the core of Coze! The visual workflow engine FlowGram is launched, and AI empowers the visual process!

ByteDance recently open-sourced its node-based process building engine: FlowGram . It is also the core visual process engine used in products such as Feishu Workflow and Button Automation.

It is a powerful tool that has been proven in online products and is extremely popular among users.

Combining node-based design and AI enhancements, it provides an intuitive visual workflow building experience. It supports  fixed layouts (structured processes)  and  free layouts (flexible connections) , automates tasks (such as data cleaning and report generation) through AI, and allows developers to extend node functionality.

It uses a front-end visual editor (React) to drag and connect nodes, and combines back-end AI integration (supports Doubao, DeepSeek, Kimi and other general large models) to provide a smooth interactive experience. FlowGram is designed to simplify workflow development while enhancing automation capabilities through AI.

It is particularly suitable for visual workflow scenarios that require clear input and output.

Core Features Highlights

• •  Dual layout mode : supports fixed layout and free layout, and can realize card-style layout (node ​​dragging) or flowchart-style layout (free line connection).
• •  Enhanced AI capabilities : It can connect to large model capabilities such as Doubao, Kimi, and DeepSeek to enhance the intelligent processing of workflows.
• •  Rich interactive experience : Provides interactive designs such as dragging, connecting, node configuration, parameter input, status highlighting, etc. to ensure clear input and output.
• •  Rich node control : branch judgment, loop execution, nested sub-processes, conditional expressions, etc. can be flexibly configured
• •  Powerful scalability : With good scalability, you can add custom functions to the node or integrate external services
• •  Low-code capability support : Possessing basic components and expansion systems for building larger-scale automation systems

Fixed layout demo:

Free layout demo:

Rapid deployment

FlowGram is developed based on TypeScript and requires the front-end NodeJS environment for local deployment.

① Install Node.js 18+

nvm install lts/hydrogen
nvm  alias  default lts/hydrogen  # set default node version
nvm use lts/hydrogen

② Clone project

git clone git@github.com:bytedance/flowgram.ai.git

③ Install global dependency packages

npm i -g pnpm@9.12.0 @microsoft/rush@5.140.0

④ Install project dependencies

Rush Update

⑤ Compile the project

Rush Build

⑥ Run the project demo

rush dev:demo-fixed-layout
rush dev:demo-free-layout

Document development: rush dev:docs (visit http://localhost:3000).
Fixed layout demo: rush dev:demo-fixed-layout.
Free layout demo: rush dev:demo-free-layout.

Create an application:

npx @flowgram.ai/create-app@latest

Last words

Traditional workflow construction often relies on a lot of coding and configuration, which is time-consuming and error-prone.  The emergence of FlowGram.AI  provides developers with a new way: through visual node editing, they can quickly build clear and intuitive workflows.

FlowGram  itself is used in ByteDance's internal products (Feishu, Coze) in a closed-source form. Now that it is open source, more companies and developer teams will use its open source features and AI-enhanced capabilities to customize efficient workflows suitable for within the company.

Its dual layout mode, rich interactive experience, powerful expansion capabilities, and AI integration plan are unique among many open source workflow tools.

Whether it is automated office, data analysis, or AI process integration, etc., it is worth integrating into your own product applications.