{"slug": "sync-all-apps-you-build-in-codex-claude-or-locally", "title": "Sync all apps you build in Codex, Claude, or locally", "summary": "Deepnote launched rebuilt data apps that let users build custom interactive apps in Codex, Claude, or local toolchains and then host, run and share them in a Deepnote workspace, with a notebook running SQL, Python or R as the backend. The apps connect to workspace data sources and native integrations including Snowflake, Databricks, BigQuery and ClickHouse, and share reusable metric definitions through notebook modules, with access controls governed by the source project's permissions. Deepnote positions the release as a shared home for AI-built apps that otherwise end up as static HTML pages with data that goes stale once exported.", "body_md": "Start in Codex, Claude, or Deepnote Cloud. Connect to your company’s data, permissions, and knowledge.\n\nToday, we’re introducing completely rebuilt Deepnote data apps: custom, interactive apps you can create in Deepnote, Claude, Codex, or your own local toolchain, then host, run and share in Deepnote.\n\nAI has made it easier for anyone to build software. Getting that software into everyday use across a company still takes work. Many vibe-coded apps end up as static HTML pages, with data that starts going stale the moment it’s exported. They lack a shared home where teammates can find them, use them, and build on them.\n\nWe’ve reimagined Deepnote apps to give them that home. Build with your preferred agent, then run and share in a workspace with live data connections, reusable metric definitions, and access controls. Your company’s data and business context inform the app from the start, and stay available as your team improves it.\n\n## **What are the new data apps**\n\nPreviously, [Deepnote apps](https://deepnote.com/docs/data-app-projects) used to be notebook blocks arranged on a page. Now, a data app has a fully custom interface, built by any agent (including the Deepnote agent) from a prompt, with a notebook as its backend.\n\n- The notebook runs SQL and Python (or, any other language, like R): queries, transformations, calculations, so your data is always fresh.\n- Notebook inputs power the app's controls: date pickers, filters, dropdowns.\n- The app is either hosted in your Deepnote workspace, next to hundreds of native integrations such as Snowflake, Databricks, BigQuery, and ClickHouse.\n\nYou can build it in Deepnote, in Claude, in Codex, or locally on your own machine. Either way, it is the same Deepnote project.\n\n## **Build from the agent you already use**\n\nStart an app in Claude, continue in Codex, or build directly with the Deepnote agent. Everyone works on the same Deepnote project, without needing the same editor or AI subscription.\n\nConnect the Deepnote plugin (via [__MCP__](https://deepnote.com/docs/deepnote-mcp) or natively in [__ChatGPT__](https://openai.com/business/plugins/deepnote/) or Claude), then ask your agent to build using your workspace’s data and existing analyses:\n\nA colleague can open that project from their preferred agent, add a forecast, or adapt it for the next leadership meeting. The app, notebook, and project files provide a shared starting point across tools.\n\n## **Keep the exploration going**\n\nDeepnote Agent suggests useful next steps based on what you’re working on. Click a suggestion to explore a new angle, compare results, or keep building your app.\n\nIf you prefer your own local setup, you can build there and publish to Deepnote when you’re ready.\n\n## **Give your apps your company’s context**\n\nAn agent can build a revenue dashboard quickly. It also needs to know what your company means by revenue.\n\nDeepnote connects agents to your workspace’s data sources, existing analyses, and reusable calculations. That context informs the app as it is built: which data to use, how a metric is calculated, and what work already exists.\n\nA revenue review and a customer health app can use the same underlying revenue definition while serving different teams. Shared notebook modules let you maintain that calculation in one place and reuse it across apps and analyses.\n\n## **Reference prior analyses in your apps**\n\nDeepnote connects agents to your workspace’s data sources, existing analyses, and reusable calculations. That context informs which data to use, how to calculate a metric, and what work to build on.\n\nA revenue review and a customer health app can share the same revenue definition while serving different teams. Define it once in a notebook module and reuse it across apps and analyses. Changes reach importing notebooks the next time they run. Agents can also reuse data from another app’s notebook, subject to the source project’s permissions.\n\nTake a look at [Lindquett Software](https://github.com/deepnote/deepnote/tree/main/examples/lindquett-software), a demo company that shares this pattern across five connected apps. The guide to [one definition of every metric](https://deepnote.com/docs/playbook-one-definition-of-every-metric) explains how to set it up and control who can change a definition.\n\n## **Fresh data with scheduled runs**\n\n[Schedule the underlying notebook](https://deepnote.com/docs/data-app-projects#scheduling) to refresh results before a weekly review or recalculate account health each morning. Give people access to the published app, and grant project access to those who need to inspect or edit it.\n\nShare an app with your team or publish it for a wider audience. Workspace controls let you manage access centrally, while the homepage helps colleagues discover apps they can use and build on.\n\n## **Control who sees what**\n\n[Give people access to the app](https://deepnote.com/docs/share-projects), the underlying project, or both. **App user** lets them use the published interface, **View** opens the project, and **Edit** lets them change it.\n\nSupported per-user OAuth integrations let queries run with each viewer’s warehouse permissions. **Private outputs** keep each person’s results separate and disables shared results.\n\n[Public apps](https://deepnote.com/docs/data-app-projects) use your workspace’s integrations. Control what they can query before sharing. Workspace admins can disable public apps entirely.\n\nSee the [Data Apps documentation](https://deepnote.com/docs/data-app-projects) for all access options.\n\n## **Start with an app your team already needs**\n\n1. In Deepnote, choose**New data app** and start with a prompt (or point to an old app you have and ask Deepnote Agent to recreate it for you in the new format).\n2. Connect via MCP to Codex or Claude and try yourself. Then, just add @Deepnote.\n3. Build Deepnote apps locally.\n\nFor inspiration, here are a [few example apps](https://github.com/deepnote/deepnote/tree/main/examples/data-apps) you can build with your Deepnote agent.\n\nWe’re excited to see what you build in Deepnote!", "url": "https://wpnews.pro/news/sync-all-apps-you-build-in-codex-claude-or-locally", "canonical_source": "https://deepnote.com/blog/apps", "published_at": "2026-11-19 15:58:50+00:00", "updated_at": "2026-11-19 16:17:21.290567+00:00", "lang": "en", "topics": ["ai-agents", "ai-tools", "agent-protocols", "ai-products", "developer-tools"], "entities": ["Deepnote", "Codex", "Claude", "Deepnote Agent", "Snowflake", "Databricks", "BigQuery", "ClickHouse"], "also_reported_by": [], "takeaway": "Deepnote's rebuilt data apps pair an agent-built custom interface with a notebook backend running SQL, Python or R, so controls like date pickers and filters map to notebook inputs and metrics defined once in a shared notebook module propagate to importing notebooks on their next run.", "recommendation": null, "alternates": {"html": "https://wpnews.pro/news/sync-all-apps-you-build-in-codex-claude-or-locally", "markdown": "https://wpnews.pro/news/sync-all-apps-you-build-in-codex-claude-or-locally.md", "text": "https://wpnews.pro/news/sync-all-apps-you-build-in-codex-claude-or-locally.txt", "jsonld": "https://wpnews.pro/news/sync-all-apps-you-build-in-codex-claude-or-locally.jsonld"}}