ByteDance open-sources Coze and Feishu workflow engines! FlowGram.AI

Written by

Caleb Hayes

Updated on:June-23rd-2025

Modern software development, data processing, and artificial intelligence (AI) operations are seeing a growing demand for complex workflow automation to manage complexity, improve efficiency, and ensure reliability. However, the challenge for developers is that traditional coding methods can be cumbersome and tedious when dealing with workflows, and standalone diagramming tools often cannot directly generate executable code. Against this backdrop, FlowGram.AI was born. This is a new project launched by tech giant ByteDance to address the above challenges. At its core, FlowGram is a node-based process building engine that helps developers quickly create workflows in either fixed layout or free-connect layout modes.

Still worried about your workflow? AI can help you free your hands!

Have you ever encountered such a situation in your daily work and study:

Want to automate a bunch of repetitive tasks but don’t know where to start?
You spent a lot of time drawing a flowchart using some diagramming software, only to find that it just “looks beautiful” but can’t actually run and solve problems?

Don’t worry, the savior is here! ByteDance recently quietly launched a new gadget - FlowGram.AI . This is also the core workflow of ByteDance’s core products COZE and Feishu.

Simply put, you can think of it as a smart "process building block box" . Just like playing with Lego when you were a kid, you only need to drag some "building blocks" (officially called "nodes") with different functions, and then connect them to quickly build a workflow.

What we are talking about today is the flowgram.ai project that ByteDance has released on GitHub. There are some similar-named projects on the market, such as flowgram.com (for codeless applications) and flow-ai.com (for AI proxy testing), so don’t confuse them!

ByteDance is not just making FlowGram.AI available for free this time for fun. They may want to create a "circle of friends" for developers around their own AI technology, allowing more people to participate in innovation, and by the way, demonstrate their technical strength and attract more outstanding talents. Moreover, the name of this tool directly contains "AI". The official statement is that AI is used . This means that AI will not only affect what your workflow can do , but also change how you do your workflow.

FlowGram.AI: How does it work?

The core of FlowGram.AI is a tool that allows you to create processes like " building blocks " and quickly complete various workflows . It is especially suitable for "visual" tasks with clear steps and a clear start and end.

It is mainly used by developers , but it also takes into account the needs of companies and enterprises and provides some advanced features, such as the ability to be deployed on the company's own servers.

The most thoughtful thing is that FlowGram.AI provides two "building modes":

Fixed Layout: Just like filling out a form, there are fixed grids for you to place "building blocks". It is especially suitable for processes with standard steps, such as processing orders and going through approvals. In this mode, it can also help you automatically generate part of the code framework, which is not easy to make mistakes.

Free Layout: This is like a blank canvas. You can place the "building blocks" as you like and connect them with various curves. It is especially suitable for scenarios that require creativity, such as designing new algorithms and building data analysis processes. It also has a very powerful technology, that is, the flowchart you draw and the actual code can be matched in real time, and the changes will be synchronized no matter which side is changed.

In fixed mode, you can also use some special "big building blocks" (compound nodes), such as setting conditions (if A is met, execute B, otherwise execute C) or loops (repeat a step). These "big building blocks" can also be nested with "small building blocks" to form a more complex process structure.

These two modes are like providing you with different styles of "building block sets". Whether it is a rigorous "standard piece" or a "creative piece" that requires free play, FlowGram.AI hopes to help you get it done. In this way, you don't have to switch between several different tools.

AI assistant is here to help: make work smarter!

The most anticipated thing about FlowGram.AI is its AI capability. It not only allows AI to perform a certain task in the process, but also integrates AI into the entire process of building and managing the process, making everything simpler and more efficient.

Intelligent Suggestions: Like a caring assistant, it will guess what functions you may need based on your current operation and recommend them to you. For example, if you have just finished "data cleaning", it may prompt you: "Do you want to try 'abnormal data detection'?"

Risk Prediction: It has a built-in smart "inspector" (machine learning model) that can help you detect up to 83% of type errors and 64% of "dead loop" risks when you test the process, and nip the problem in the bud in advance!

Automated Documentation: Every time you save your built process, it can automatically generate a technical document containing flowcharts, interface descriptions and test cases. It is said that this is 5 times faster than writing it manually! No more headaches for document writing!

These AI functions are not just for show, they are real "magic assistants" that help you improve efficiency and reduce errors. "Smart recommendations" broaden your thinking, "risk warnings" help you control quality, and "automatic document writing" liberates productivity. This deep AI integration is a highlight that distinguishes FlowGram.AI from many other tools.

A little technical knowledge: what is different about FlowGram.AI?

FlowGram.AI uses a lot of cutting-edge technologies to ensure that it is very smooth to use, and can easily handle even very complex flowcharts. Its design inspiration even borrows some ideas from game engines, with the goal of making the interactive experience better.

It mainly uses a modern programming language called TypeScript (which can be understood as a more advanced and less error-prone JavaScript). Moreover, it can also be used well with popular web development technologies such as React and Vue.

In order to ensure that you can use it without any lag, it also uses some optimization methods, such as putting some complex computing tasks into the background to ensure the smooth operation of the main interface; only updating the changed parts of the interface instead of refreshing the entire interface; and caching some commonly used things to speed up loading.

Simply put, FlowGram.AI is more like a new generation of tool tailored for developers who are familiar with modern web technologies, deeply integrated with AI capabilities, and capable of visually building various workflows (even including some applications with user interfaces).

FlowGram.AI installation and configuration (technical students see here)

If you are a technology enthusiast or developer and want to experience FlowGram.AI yourself, the official also provides a guide.

# create demo
npx @flowgram .ai/create-app @latest

# in PowerShell
npx  " @flowgram .ai/create-app @latest "

# select demo
- fixed-layout  # full-feature overview
- free-layout  # full-feature overview
- fixed-layout-simple  # basic usage
- free-layout-simple  # basic usage

You need to prepare a Node.js 18 or higher environment, as well as the two tools PNPM and Rush. The specific installation steps can be found in its GitHub project page.

There are also detailed tutorials in the official documentation, teaching you step by step how to configure the canvas, define data, declare "building blocks", etc.

GitHub repository: github.com/bytedance/flowgram.ai is the "base camp" of FlowGram.AI, where all codes, problem discussions, and examples are located.

Official documentation: The official documentation is on flowgram.ai, but the website currently has relatively little content. More detailed technical documentation and tutorials are actually stored in the form of files in the apps/docs/src/en/ directory of the GitHub repository.

For those who are new to the project, it may be a bit confusing to find the documentation, but don’t worry, the important information is all in the GitHub repository.

The power of open source: use it for free and contribute together!

FlowGram.AI is an open source project , the code is on GitHub, and ByteDance contributed the most. It uses the MIT license , which is a very permissive license, meaning that you can use it for free, modify it, and even use it in commercial projects, as long as you keep the original copyright and license statement. This open attitude will help attract more people to use and participate in improving FlowGram.AI.

GitHub Issues (Problem Feedback Area): You can see the questions raised and the bugs reported here, and the developers will also respond here.

GitHub Discussions: This is where you can have more extensive conversations and share ideas.

Feishu Group: Join the official Feishu group to communicate directly with the development team and other users.

Last words: Will FlowGram.AI be your next efficiency tool?

In general, FlowGram.AI, as a new attempt by ByteDance in the field of workflow automation, does have many eye-catching aspects.

Its biggest advantage lies in its flexible construction method and in-depth AI capabilities . It not only lets AI do your work, but also lets AI help you better "design" how to do your work. Coupled with its fashionable technology selection and smooth interactive experience , as well as its open source and free features and the support of large companies , FlowGram.AI has great potential for future development.

For ordinary users like us, although we may not directly build very complex processes, understanding such tools can let us see how AI technology can actually change the way of working and improve efficiency. Perhaps in the future, many tedious processes in our lives and work can be made simpler and smarter through tools like FlowGram.AI.

If you are interested in this type of AI efficiency-enhancing tool, you might want to pay attention to the subsequent development of FlowGram.AI. Let's see what surprises this "smart building block box" will bring us in the future! Maybe it can really become a "efficiency artifact" that frees many people's hands.