Lucidchart export to mermaid, by redrawing the decision

A lucidchart export to mermaid is a redraw. Lucidchart's export is built around images and Visio-style files. I am not going to promise a native Mermaid download, and I am not going to describe a button I cannot stand behind. If your own account shows an export this page does not mention, believe the menu in front of you. For the accounts and the help pages I can talk about honestly, the way out of a Lucidchart file and into a text diagram is to pick one decision flow, redraw it, keep the Lucidchart file as the picture spec, and stop editing both once the text is the source.

The Mermaid versus Lucidchart comparison is the longer argument about workshops, seats, and which tool should own which picture. This page is the migration of a single chart. I would rather finish one flowchart than announce a program to convert the company drive.

What the export menu is for

Lucidchart is a canvas. People drag shapes, comment, and leave with a picture that can look like a slide. That is a real product, and it is the right room for a workshop where the layout is the conversation. When the workshop ends, the file you can take out is typically an image, or a Visio-style file for a tool that thinks in Visio. Those outputs are pictures or foreign canvases. They are not a fence you can paste into a README and diff next month.

I keep saying "Visio-style" on purpose. I am not going to invent which plan includes which Visio variant, or quote a price. Prices and packaging move, and a number I typed from memory would be a rumor with a dollar sign. Open Lucidchart's own pricing if you are buying. Open the export menu on your file if you are leaving. The menu is the spec for your account. This page is the spec for what to do when Mermaid is not on it.

An image export is still useful. It is the picture spec. Print it, or park it beside the editor, and rebuild the decisions in text. Do not trace pixels. Trace the questions. A diamond in the image becomes a diamond in Mermaid. A decorative icon becomes a word, or it gets cut. The redraw is allowed to be uglier. It has to be editable.

The draw.io conversion guide is a different job. That guide exists because a draw.io file can be inspected as a document you can read. I am not going to pretend a Lucidchart export hands you the same kind of source. If you already landed in draw.io from somewhere else, use that guide. If you are holding a PNG and a .vsdx-style file from Lucidchart, you are holding a picture and another canvas. Redraw.

Pick one decision flow

Do not migrate the archive. An archive of org charts, journey maps, and posters will eat a quarter, and most of those pictures are not hurting you as images. Pick the diagram that changes when the code changes, or the one a review keeps getting wrong. A checkout decision, an approval threshold, a deploy rollback. One flow, one owner, one place it will live after the redraw.

I choose by asking which picture I was afraid to edit last month because I did not have the canvas open. That fear is the signal the source is in the wrong tool. A diagram nobody edits can stay an image forever. A diagram that should have changed in the same pull request as a handler should be text.

Name the scope in a sentence before you draw. "Refunds under the threshold skip finance. Refunds over the threshold wait." If you cannot write the sentence, you are not ready to redraw. You are ready to look at the Lucidchart file and argue about what it meant. Have the argument. Then write the sentence. The sentence is the spec. The old canvas is evidence. I have redrawn a chart faithfully, including the ambiguous diamond, and thereby preserved a bug in a more diffable format. Faithful and wrong is still wrong. Resolve the ambiguity in words, then draw the words.

Leave every other Lucidchart file where it is. A migration plan that starts with "export all" is how the one flowchart you needed stays unconverted while you rename folders. One file this week. The next file when the next review hurts.

Redraw from the decisions, not from the coordinates

Open the image. List the decisions and the end states. Ignore fonts, colors, and the exact x and y of each box. Mermaid will lay the picture out. You will not recover the canvas, and chasing it is how the redraw stalls. The flowchart syntax reference is the grammar for the new file: a node id, a label, an arrow, a label on the arrow when the arrow is a yes or a no.

Ids are short handles. Labels are the words from the sentence. Do not paste a paragraph into a node because the Lucidchart box was wide enough to hold one. The box was wide because someone dragged it. The label should survive a narrow README column. Put the paragraph under the figure, in the same Markdown file as the fence.

Here is a first redraw of a refund approval I have seen on canvases, including the mess the canvas usually has: a start oval, a vague "process refund" box, and a decision that mixes the amount with a manager's mood.

flowchart TD
  startNode[Request arrives] --> vibe{"Feels ok?"}
  vibe -->|Yes| magic[Process the refund]
  vibe -->|No| hold[Hold for someone]
  magic --> doneBox[Customer is paid]
  hold --> doneBox

That chart is honest about a bad source. "Feels ok?" is not a policy. I do not ship it. I show it because the first redraw often copies the ambiguity, and you should notice that before you commit. The Lucidchart file remains the picture spec while you interrogate it. Ask the owner what "feels ok" meant in the meeting. Write down the answer. Then draw the answer, and only the answer.

The second flowchart is the one I would commit. The threshold is a number the sentence owns. The hold has a destination. The paid state has a name that is not end, because end is a Mermaid keyword and a future subgraph would swallow the file.

flowchart TD
  request[Refund request] --> limit{"Under the limit?"}
  limit -->|Yes| pay[Send the refund]
  limit -->|No| finance[Wait for finance]
  finance --> approved{"Finance approves?"}
  approved -->|Yes| pay
  approved -->|No| tell[Tell the customer no]
  pay --> paid[Record the payout]
  tell --> closed[Record the refusal]

The image from Lucidchart can sit in the ticket until this fence is reviewed. Reviewers who need the old layout can look at the image. Reviewers who need the policy can read the fence. When the fence is accepted, the image becomes a snapshot with a date, not a second editor. I write that date in the commit message so nobody "syncs" the canvas back a month later and calls it a fix.

Stop maintaining both

Two sources feel safe and behave like a bug. Someone will update the canvas during a workshop and forget the README. Someone else will update the README during a pull request and forget the canvas. The next incident will quote whichever picture looks more official, and the polished one usually wins, even when it is older. Polished is not current.

Pick the text once the redraw has been reviewed. Say so at the top of the doc: the fence is the source, the Lucidchart file is a snapshot from a date you actually type. Then stop opening the canvas for this flow. If a workshop needs a canvas again, export a fresh image from the text, or redraw on the canvas for the hour and throw that board away. Do not quietly promote the board back to source because it was easier to drag.

I keep the old file for a while, read-only, because people panic when a link dies. Read-only is a permission, or it is a sentence that says "do not edit." A sentence is weaker and still worth writing. Delete the file later, when the fence has survived a few real changes. Deleting it on day one creates a scavenger hunt. Leaving it editable creates a fork.

Visio-style exports do not change this ending. A .vsdx you can open in another canvas is still not Mermaid. It can be the archive copy. It should not be the copy the pipeline review updates. The Visio flowchart maker note is for people who are still choosing that canvas. The draw.io Mermaid support note is the neighboring migration, the one with a file you can read. Neither one adds a Mermaid item to Lucidchart's export menu.

What I would leave on the canvas

Leave the workshop artifacts. A room full of stickies, a customer journey with illustrations, a slide that took an hour to align. Mermaid will make those look like a spec, because it is a spec language. Stakeholders who wanted a designed leave-behind will say so, and they will be right. Export the image from Lucidchart for that audience. Do not punish a poster by forcing it through a renderer that does automatic layout.

Leave anything you do not understand yet. A redraw of a mystery chart spreads the mystery into Git, where it looks decided. Park the image in the ticket with questions on it. Redraw after the questions have answers. I have committed "TODO" diamonds that lived for a year because they rendered cleanly. A clean render is a poor excuse.

Leave charts whose editors will not open a fence. If the only people who maintain the picture will not edit Markdown, the text file will rot, and the canvas was the honest tool. The flowchart online tool and free diagram tool notes are about that choice from the Mermaid side. Use them when you are deciding where new work goes. Use a redraw only when the maintainer of this chart will actually maintain text. A migration that ignores the maintainer is a screenshot with extra steps.

The AI flowchart generator can take your sentence, the one you wrote before redrawing, and emit a first fence. Free accounts include 5 AI uses in total, and the same panel can edit or fix the result. Paste the sentence, not a description of Lucidchart's colors. Then delete every node the sentence does not need. I treat the generator as a typing assistant. The picture spec still wins on meaning, and the sentence wins on scope. If the generator adds a fraud vendor nobody mentioned, remove the vendor. Extra lifelike detail is how the old canvas's clutter moves into Git and starts looking intentional.

A small checklist for the one file

I keep this next to the image while I redraw, and I throw it away after.

1. One sentence that states the decision, including the unhappy path.
2. A list of diamonds, each of them a question with yes and no.
3. A list of end states with ids that are not end.
4. A fence in the editor, previewed live, labels short enough for a README.
5. A review against the image, looking for a decision the fence dropped or a decision the image never really made.
6. A line in the doc that names the fence as the source and the Lucidchart file as a dated snapshot.
7. The canvas left alone on the next change. The next change is a text edit.

Step 5 is the one I try to skip when I am tired. It is the step that catches "feels ok." Compare the fence to the image and to the sentence. The image is allowed to lose. The sentence is not, unless you change the sentence in the open and tell the owner.

If the fence and the image disagree and you cannot find the owner, do not publish either as policy. Publish the question. A diagram tool is not a substitute for the person who remembers the meeting. The comparison with Lucidchart is where I send someone who is still choosing a tool for the next workshop. Send them this page only when the workshop already happened and the file needs a life in Git.

Draw the sentence, then close the canvas

Put the image where you can see it. Write the decision in a sentence. Draw that sentence as a flowchart with boring ids and short labels. Review it against the image once. Name the text as the source. Stop updating the canvas for that flow.

Open the editor and paste the second refund chart if you want a starting fence, or paste your own sentence into the generator and then delete the extras. The editor is free without an account, and the preview is live, which is the property you want while you are still finding the real diamond. The Lucidchart file can keep the colors. The next edit belongs to the fence.