A reactive notebook for Python — run reproducible experiments, query with SQL, execute as a script, deploy as an app, and version with git. Stored as pure Python. All in a modern, AI-native editor.
Python artificial-intelligencedagdata-sciencedata-visualizationdataflowdeveloper-toolsmachine-learningnotebookspipelinepythonreactivesqlweb-app
/* Copyright 2026 Marimo. All rights reserved. */
import { hasRunAnyCellAtom } from "@/components/editor/cell/useRunCells";
import { autoInstantiateAtom } from "@/core/config/config";
import { getInitialAppMode } from "@/core/mode";
import { store } from "@/core/state/jotai";
export interface MarimoExportContext {
trusted: true;
notebookCode?: string;
}
declare global {
interface Window {
__MARIMO_EXPORT_CONTEXT__?: Readonly<MarimoExportContext>;
}
}
function isMarimoExportContext(
value: unknown,
): value is Readonly<MarimoExportContext> {
if (typeof value !== "object" || value === null) {
return false;
}
const candidate = value as MarimoExportContext;
if (candidate.trusted !== true) {
return false;
}
if (
candidate.notebookCode !== undefined &&
typeof candidate.notebookCode !== "string"
) {
return false;
}
return true;
}
export function getMarimoExportContext():
| Readonly<MarimoExportContext>
| undefined {
const context = window?.__MARIMO_EXPORT_CONTEXT__;
return isMarimoExportContext(context) ? context : undefined;
}
export function hasTrustedExportContext(): boolean {
return getMarimoExportContext()?.trusted === true;
}
/**
* True when the current page is a context where notebook-authored script
* execution is expected, and therefore the user has consented (explicitly or
* by the nature of the page) to running arbitrary notebook content:
*
* - the user has run at least one cell, OR
* - a first-party exported notebook page installed a trusted export context
* (islands / static exports / Quarto islands), OR
* - `auto_instantiate` is enabled (the notebook runs on page load by user
* configuration), OR
* - the page was loaded in `read` / app mode (served by marimo as an app).
*
* Edit mode before any user interaction is intentionally NOT trusted — that
* is the only surface where we must prevent notebook-authored content from
* loading scripts or bypassing HTML sanitization.
*/
export function hasTrustedNotebookContext(): boolean {
if (store.get(hasRunAnyCellAtom)) {
return true;
}
if (hasTrustedExportContext()) {
return true;
}
if (store.get(autoInstantiateAtom)) {
return true;
}
try {
if (getInitialAppMode() === "read") {
return true;
}
} catch {
// getInitialAppMode throws before mount config is applied; treat as untrusted.
}
return false;
}