Navigate and preserve state

Each named view has a stable route. The configured default opens at /. A view named report opens at /report/.

Studio keeps public query parameters and the URL fragment with the active presentation while removing private routing and credential parameters from notebook query state.

Link named views

Use sibling-relative links when a group of views should work under a live server, a deployment base path, and a copied static export:

<nav aria-label="Analysis views">
  <a href="../dashboard/index.html">Dashboard</a>
  <a href="../report/index.html">Report</a>
</nav>

Export linked views into sibling directories:

dist/analysis/
  dashboard/
    index.html
  report/
    index.html

Studio carries the active public query and fragment through an in-product view switch. An authored link controls its own destination URL.

Read and write public query parameters

Use mo.query_params() in the notebook:

@app.cell
def filters(mo):
    query = mo.query_params()
    region = query.get("region", "all")
    return (region,)

Open the view with a public parameter:

/report/?region=emea

Changes made through mo.query_params() synchronize with Preview and browser history. Repeated keys and blank values remain distinct:

/?tag=first&tag=second&empty=

Studio keeps routing keys such as runtime, session_id, access tokens, and its marimo_studio_* capabilities out of the public notebook query.

Use fragments for local navigation

An anchor stays inside the current view:

<a href="#error-evidence">Review errors</a>

<section id="error-evidence">...</section>

Studio applies the fragment after the presentation is ready. Browser back and forward navigation restores the public query, fragment, and selected view.

Choose runtime from the URL

When both runtimes are configured, runtime selects the notebook runtime for the request:

/?runtime=wasm
/?runtime=server

The parameter must name one of the configured runtimes. Studio treats it as routing state, so notebook code does not receive it through mo.query_params().

Open multiple tabs

In edit mode, Studio uses Marimo's native sessions. Tabs opened on the same notebook and server share one Python kernel. Changing a control recomputes notebook outputs across Python previews. Widget state and public query parameters synchronize across those tabs. Each Studio tab keeps its selected view.

Marimo gives the first connection editing control. Other tabs can interact with the notebook through its kiosk view. Select Take over in the embedded editor to move editing control to that tab. Closing the editor tab leaves the remaining tabs connected to the same kernel.

Run mode follows Marimo's visitor sessions. The Browser runtime runs a separate notebook in each browser worker. Selecting another runtime changes which notebook instance supplies that tab's outputs.

Preserve a Python session across navigation

Enable session replay in the notebook configuration:

[tool.marimo-studio]
runtime = "server"
runtimes = ["server", "wasm"]
preserve_session = true

In run mode, Studio can reconnect a navigation or reload to the existing Python runtime session when all of these still match:

• saved notebook
• Python runtime
• view path and public query
• current server process and live session

A different public query starts or selects the session for that query. A missing, expired, or mismatched session starts a fresh one. The Browser runtime manages its notebook instance in the visitor's browser and does not use preserve_session.

Leave preserve_session = false when each reload should receive a fresh Python session. Continue with Run or export a view to choose a runtime and delivery path.