Link views and keep state ​

A notebook's views can work as one site. Each view has a stable URL, so a dashboard can link to its report, and a link can carry query parameters that the notebook reads, so the report opens filtered to the region the reader picked. The default view opens at /, and a view named report opens at /report/.

Use sibling-relative links, so the links keep working under a live server, a deployment base path, and a copied static export:

html
<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 folders, such as dist/analysis/dashboard/ and dist/analysis/report/.

Share state in the URL ​

Read and write query parameters in the notebook with mo.query_params():

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

/report/?region=emea then opens the report filtered to EMEA. Changes the notebook makes through mo.query_params() update the browser URL and history. Repeated keys and blank values stay distinct, as in ?tag=a&tag=b&empty=.

An anchor such as <a href="#errors"> stays inside the current view. Studio applies the fragment once the view is ready, and back and forward restore the query, the fragment, and the selected view. Switching views in Studio carries the current query and fragment along.

Studio keeps its own parameters, such as runtime, session_id, and access tokens, out of mo.query_params(). When a notebook allows both runtimes, ?runtime=wasm or ?runtime=server picks one for the request.

Open multiple tabs ​

In marimo edit, tabs on the same notebook share one Python kernel. Changing a control in one tab recomputes results in every Python preview, and widget state and query parameters sync across tabs. Each tab keeps its own selected view.

The first tab has editing control. Other tabs can still use the notebook, and Take over in the embedded editor moves editing control to them.

In marimo run, each visitor gets their own session. The Browser runtime runs a separate notebook in each browser.

Keep a Python session across reloads ​

By default, a reload in marimo run starts a fresh Python session. Turn on preserve_session to reconnect to the existing one:

toml
[tool.marimo-studio]
preserve_session = true

Studio reconnects when the saved notebook, the Python runtime, the view path and query, and the server process all still match. Anything else starts a new session, and a different query selects the session for that query. The Browser runtime ignores this setting.