Connected repositories
The code is the only description that is never out of date.
So that is where the docs and diagrams come from. Connect a repository once, link a document to it, and the description of your system is rewritten from the code whenever you ask.
- Read-only repo access
- Works with Claude Code, Cursor, Claude Desktop
- Free for one editor

What it writes
Five kinds of document, each with its diagram.
One repository feeds as many documents as you need. Each is linked to the repo on its own, with its own kind and its own instructions, and refreshes on its own.
High level architecture
Data lineage
Security review
Solution architecture
User guide
Your instructions, kept

Every one of them is a real DocZoid document with a real DocZoid diagram. Real containers, real edges on real anchor points, a canvas you can drag things around on. Not a rendered image, and not a Mermaid file that you now have to find a home for.
Refresh
Three ways to bring it up to date.
A refresh reads the current code and rewrites the document and its diagram in place. You decide when: by hand, from your AI client, or from CI on every push.
A refresh does not throw away your work. A version snapshot is taken first, your instructions to the AI are applied again, and the result is written into the live document, so people with it open see the change arrive. When a service appears or disappears in the repository, it appears or disappears in the diagram, and every PortalSnip of that diagram updates in place.
How PortalSnips work
- In the editorEvery repo-linked doc carries a Refresh with AI button. A version snapshot is taken before the run, so a bad refresh is one click to undo.
- From your AI clientThe MCP server exposes the same refresh as a tool, so Claude Code or Cursor can refresh the doc for the change it just made.
- From CI, on every pushThe refresh is an authenticated API call. A workflow step on main runs it on every push, and the doc never waits for a person.
Access
What DocZoid can see, and what it will never do.
Read-only, always
You choose the repositories
Bring your own model
Or keep it local

Hosts
Where your code lives.
Need one that is not on the list? Tell me and it moves up the queue.
Not only from code
You can also just describe it.
The same generator works on a paragraph. Write what the system does and DocZoid lays out the diagram beside it, on the canvas, ready to edit.
Questions
The ones people ask first.
- Does DocZoid write to my repository?
- No. Access is read-only. DocZoid reads your code to understand the system and never pushes a commit, opens a pull request, or changes a file.
- Can I edit a generated diagram?
- Yes. A generated document is an ordinary DocZoid document: edit the prose, move a box, restyle a container, add a note. A refresh rewrites it from the current code, after taking a version snapshot, so you can always step back.
- What about private repositories?
- Supported. You choose which repositories DocZoid can see when you connect it.
- Which model does the generating?
- You can bring your own key, or point DocZoid at a local model over MCP if your code is not allowed to leave your network.
- What happens to a monorepo?
- Connect the repository and scope a diagram to a directory. One repo can feed several diagrams, each with its own scope, in different documents.
- Which plan includes this?
- Connected repositories are included from the Team plan.
Connect a repo. See your architecture in five minutes.
- Read-only repo access
- Works with Claude Code, Cursor, Claude Desktop
- Free for one editor