Use Your AI Agent to Create and Share Mermaid Diagrams in Seconds

Most diagramming tools treat AI as a feature inside their editor — you open the app, click an AI button, type a prompt, and get a diagram. It works, but it’s manual. You’re still context-switching away from whatever you were doing to go make a diagram.

What if your agent could just do it for you? You’re in Claude Code reviewing an auth flow, and you say “diagram this and share it with the team.” Thirty seconds later, a live diagram is hosted at a shareable URL, and you never left your terminal.

That’s how MMD Studio works with AI agents.

Why agents and Mermaid are a natural fit

Mermaid is already the most AI-friendly diagramming format. It’s plain text, well-documented, and every major LLM can write it fluently. When you ask Claude or GPT to diagram something, they reach for Mermaid syntax by default — because it’s the format they know best.

The missing piece has always been what happens after the agent writes the Mermaid code. Typically, you get a code block in your chat. You copy it, paste it into an editor, render it, export it, and share it manually. The agent did 10% of the work and you did the other 90%.

MMD Studio closes that loop. The agent writes the Mermaid code, creates the diagram via the API, sets it to shareable, and hands you a link. One prompt, one URL, done.

How it works

MMD Studio exposes two things that make agent integration trivial:

A REST API that supports full CRUD on your diagrams — create, read, update, delete, and control sharing settings. Every action you can take in the browser, the API can do programmatically. Authenticate with a bearer token and you’re off.

A public skill file at mmd.studio/SKILL.md that teaches any agent the API contract in one fetch. The agent reads the skill, understands the endpoints, and starts making calls. No plugin to install, no OAuth flow to configure, no marketplace to browse.

What this looks like in practice

Here are a few real prompts you can hand to an agent with your MMD Studio token:

“Diagram the signup flow and share it with the team.” The agent writes a flowchart (landing page → email form → verify → dashboard), creates it on MMD Studio via the API, flips visibility to “anyone with the link,” and returns a mmd.studio/d/... URL. Paste it in Slack and your team sees a live, always-current diagram.

“Read our auth diagram and add a refresh token step.” The agent fetches the existing diagram, parses the Mermaid source, adds the new step, and pushes the update. Anyone with the existing link sees the updated version immediately — no re-sharing needed.

“List all my diagrams and tell me which ones are shared.” The agent calls the list endpoint and gives you a summary. Useful for auditing what’s public and what’s private.

“Here’s a share link — summarize what this diagram shows.” The agent fetches the public diagram, reads the Mermaid source, and explains it in plain language. Great for onboarding or reviewing someone else’s architecture docs.

Setting it up

You need two things: an MMD Studio account (free works, Pro removes API rate limits) and an API token.

1. Sign in at mmd.studio
2. Go to Settings → API tokens
3. Create a token and save it as an environment variable: export MMD_TOKEN="mmd_..."

From there, point your agent at the skill file and start prompting.

Claude Code — the skill file works natively. You can install it locally so it auto-loads every session:

mkdir -p ~/.claude/skills/mmd-studio
curl -sS https://mmd.studio/SKILL.md -o ~/.claude/skills/mmd-studio/SKILL.md

Cursor, Claude Desktop, or any agent with web fetch — just tell it to read the skill URL at the start of the conversation:

Read https://mmd.studio/SKILL.md and follow those instructions.
My MMD_TOKEN is mmd_...

That’s the entire setup. No SDK, no dependency, no plugin.

Why this matters for teams

The value of agent-created diagrams isn’t just speed — it’s that diagrams actually get made. Most architecture decisions, onboarding flows, and system designs should have a diagram but don’t, because the friction of creating and sharing one is just high enough that people skip it.

When creating a diagram is a single sentence in your agent prompt, the calculus changes. Diagramming stops being a task you schedule and becomes something that happens as a side effect of working. Your team’s diagram library grows organically because the cost of adding to it is nearly zero.

And because MMD Studio diagrams are hosted and shareable by default, every diagram the agent creates is immediately a living reference document. Drop the link in a PR description, pin it in a Slack channel, or add it to your wiki. It stays current because updating it is another one-line prompt.

Compared to in-app AI features

Some diagramming tools offer AI as a built-in feature — you get a chat interface inside the editor that generates diagrams from natural language. That works fine if you’re already in the editor and want to stay there.

But it means you’re locked into one workflow: open the app, use their AI, stay in their UI. You can’t trigger it from your terminal, your CI pipeline, your code review tool, or your own agent. You can’t script it. You can’t automate it. And you’re typically paying for AI credits on top of your subscription.

MMD Studio takes the opposite approach. The API is open, the skill file is public, and any agent that can make HTTP requests can drive the entire platform. Your AI, your workflow, your tools — MMD Studio just provides the hosting and sharing layer.

Get started

1. Open mmd.studio and create a free account
2. Generate an API token in Settings
3. Point your agent at mmd.studio/SKILL.md
4. Ask it to create your first diagram and share it

The whole process takes about two minutes, and from then on, every diagram is one prompt away.

MMD Studio is a fast, clean Mermaid diagram editor with a REST API and agent integration built in. Free to start, $30/user/year for unlimited diagrams and API requests.