
Mermaid
The visual language every AI already speaks
4.8•27 reviews•1.4K followers
The visual language every AI already speaks
4.8•27 reviews•1.4K followers
Mermaid diagrams are text, so they version with your code, travel with your PRs, and can be read, generated, and acted on by every major AI platform, including Claude Code, Cursor, ChatGPT, Gemini, and GitHub. 200k+ teams use Mermaid to keep complex systems legible, for the humans building them and the agents working alongside. Built on Mermaid.js, the open-source standard with 90k+ GitHub stars.
This is the 4th launch from Mermaid. View more

Mermaid Flow
Launching today
Mermaid Flow lets you build agentic workflows as diagrams, then run them.
Think issue triage, weekly recaps, PR reviews: work you'd otherwise hand to a long prompt or a skill file. Describe what you want, start from a template, or bring the file, and you get a workflow with every step laid out. Connect Github, Slack, Notion, and more in one click, then hit run or put it on a schedule. Every run is saved step by step, so you can see what happened and your team can follow along. Free to start.





Free Options
Launch Team

CotypistType up to 50% less in every Mac app. Free to try, and it runs fully on your Mac.
Promoted





Having seen Mermaid Flow grow from the very beginning makes this launch especially meaningful to me! It's amazing to see how far we've come and what we've built together.
What I love most is how it brings together visual editing and code, letting you build and run AI workflows in whichever way feels most natural. The diagram isn't just a visualization, it's the actual executable workflow!
Really proud to be part of this journey.
Excited for what's next! 🚀
@gabry_squared Onwards and upwards! Appreciate all the help shaping this! 🎉
Congrats on the launch everyone… there are many workflow tools out there, but many don't give you predictability and transparency into how it's working, and none of them are built on the open source foundations of Mermaid. We built a whole new diagram syntax that's specifically for agents and is human readable. People have been building with Mermaid-js for over a decade, and Mermaid is as important as ever, now diagrams are not only outputs they're inputs.
@nimprojects Well said and pleasure working together on this! ✨
I already keep Mermaid diagrams in repos and PRs, so the part that clicks for me is that the diagram is the workflow, not a picture of it — it versions with the code instead of going stale in a doc somewhere. The deterministic steps point matters too: the usual failure mode I hit with agents is them deciding to skip or reorder steps. When a step produces a bad output mid-run, can you rerun just that step, or does the whole workflow start over?
@jcbuilds Great to hear the approach resonates! 🎉 I've also found that mixing deterministic steps with agent reasoning is a beautiful combination. On your question: Today, the whole workflow will re-run, but we're actively looking into adding the ability to stop at a given step and check the outputs, then continue from there. 💯
So proud of this team pulling together a major launch that I think many Mermaid enthusiasts will love! You keep the Open source syntax in the background, but pull in the modern agentic workflows that customers are sprinting towards today. I'm very excited to see how everyone puts Flow into play!
@dominic_dewalt Thanks for sharing more context! Spot on 🚀
As a data person, I love how Mermaid Flow simplifies the recurring work that eats up my week: pulling KPIs, checking the dashboards, and flagging what moved. I described those jobs in plain English, and in about 2 prompts it built flows that pull from Hex, Mixpanel and CRM tools and send a summary to my Slack. Every flow is a readable diagram, so I can see exactly what each step does and explain it to my team.
@yuki_liu Great use cases! Agents can do amazing data work with the right structure, especially when we have visibility into the logic behind the scenes. 🤖
our inbox triage runs on an agent every morning and the prompt only keeps growing, so that's the one I'd move over. can a step wait for a person to approve before the run goes on?
@olia_nemirovski Triage is a great use case! For ever-growing prompts, it should indeed be a lot easier to iterate with your team when it's visualized on a diagram with clear steps.
On approvals: Human-in-the-loop is on our short list of next improvements. 🔥 In the meantime what I've found works is ending the workflow with a doc, pull request, or DM to me in Slack so that I can review it before hitting "Send." Would love to hear any feedback here as you dig in since this is towards the top of our list. 🙌