# Streamlit: the agent-native data app framework for Python

Source: https://streamlit.io/

The agent-native data app framework

## Built in Python. *Made for people and agents.*

Turn your data into interactive apps and dashboards, written in plain Python by you or your coding agent. People get fast, polished apps, and agents use the same apps as tools, with the same numbers and access rules.

[Get started](https://streamlit.io/#get-started) [Set up your coding agent](https://streamlit.io/agents#skills) [Try the demos](https://streamlit.io/#use-cases)

`pip install streamlit` [Open source](https://github.com/streamlit/streamlit)

app.py Plain Python

```
import streamlit as st
from metrics import REGIONS, kpis, by_month

st.title("Net revenue")
st.caption("Excludes refunds. Amounts in EUR.")

with st.container(horizontal=True):
    region = st.selectbox("Region", REGIONS)
    quarter = st.segmented_control(
        "Quarter", ["Q1", "Q2", "Q3", "Q4"],
        default="Q3", required=True,
    )

k = kpis(region, quarter)
grid = st.grid(3, border=True)
grid.metric("Revenue", k.revenue, k.revenue_delta)
grid.metric("Orders", k.orders, k.orders_delta)
grid.metric("Avg. order", k.aov, k.aov_delta)

st.bar_chart(by_month(region), x="month")
```

One app, two audiences

Net revenue

Excludes refunds. Amounts in EUR.

Region

Quarter

Revenue

€4.11M

↓ 13.8%

Orders

18.4k

↑ 2.2%

Avg. order

€223

↓ 15.7%

No frontend, API layer, or contract to write. Same logic and access rules for people and agents.

## The No. 1 Python framework for data apps and dashboards

Millions of people use it, at thousands of companies, from two-person startups to global enterprises.

GitHub repositories depend on it[GitHub dependents page, November 2026](https://github.com/streamlit/streamlit/network/dependents)

1M+

Library for creating dashboards[Python Developers Survey 2024 by the PSF and JetBrains](https://lp.jetbrains.com/python-developers-survey-2024/)

No. 1

Downloads in the last month[PyPI via pypistats.org, November 2026](https://pypistats.org/packages/streamlit)

23M+

GitHub stars[GitHub, November 2026](https://github.com/streamlit/streamlit)

46k

What teams build

## Dashboards, forecasting, and workflow apps

A sales manager sees margin fall on a dashboard, asks an agent to compare three price scenarios with Finance's forecasting app, and approves the change in a workflow app. Each is a live Streamlit app, built from the same polished components, with theming and agent access built in.

- [Live demo](https://advanced-dashboard-demo.streamlit.app/)

  Dashboards

  ### KPI dashboards with drill-down and cross-filtering

  Product metrics, operating views, and live monitoring over large tables, on any data source. Built with your coding agent, on logic your team can read.

  [See dashboards →](https://streamlit.io/dashboards)
- [Live demo](https://forecasting-app-demo.streamlit.app/)

  Forecasting apps

  ### Simulations and models with what-if levers

  Scenario planning, pricing and demand simulators, and churn and risk models. The computation and the interface are the same Python.

  [See forecasting apps →](https://streamlit.io/use-cases#forecasting)
- [Live demo](https://workflow-app-demo.streamlit.app/)

  Workflow apps

  ### Review queues, approvals, sign-offs, and writeback

  Including approving what an AI agent proposes. The decision happens in the same app as the data, checked by the same Python logic that produced the numbers.

  [See workflow apps →](https://streamlit.io/use-cases#workflows)

Streamlit is also great for [AI chat and agent apps over your data](https://streamlit.io/generative-ai), ML and model tools, and internal data tools.

Why Streamlit

## Generating an app is cheap. *Owning it isn't.*

An agent can generate a React frontend and a Python API in minutes. Then you own both, the contract between them, a few hundred frontend packages, and numbers that have to stay right as the logic changes.

Streamlit apps are plain Python your data team can read, with a consistent, accessible design and no frontend or API layer to write. Every upgrade improves all of your apps at once, security fixes included, so a hundred apps don't mean a hundred upgrade projects.

Developers write any logic they need. Security and platform teams review the framework once, so each app's own review stays small.

### When a custom stack is the better choice

- The UI itself is the product
- The app is customer-facing at large scale
- It must be mobile-first or work offline
- It needs a design that Streamlit's theming and custom components can't reach

[See the full comparison](https://streamlit.io/why-streamlit#compare)

Streamlit

Python

- app.py
- metrics.py
- .streamlit/
- config.toml

- Batteries included, no npm tree
- Authentication, caching, and state built in
- Already a tool for agents

A custom stack

TypeScript and Python

- web/
- package.json
- src/pages/revenue.tsx
- src/components/KpiCard.tsx
- src/components/RevenueChart.tsx
- src/hooks/useRevenue.ts
- src/api/client.ts
- api/
- main.py
- routes/revenue.py
- schemas.py
- auth.py
- cache.py

- Two toolchains and a large npm tree
- Authentication, caching, and state assembled per app
- A second API for agents

Agent skills, built in

## Built for coding agents

Every release ships with expert skills for coding agents, so they build Streamlit apps that look good and run fast. We write them ourselves, and they match the version you installed, not the one in your agent's training data.

- ### One command

  Installs the skills for Claude Code and for agents that read the shared `.agents` folder. Then ask for a Streamlit app.
- ### Less code to own, fewer places to break

  No frontend, API contract, or build setup to generate. Compared with a custom stack, each app has less to maintain and fewer places to break, and it takes fewer tokens to write and change.
- ### Agents verify their own work

  They drive the running app without a browser and read its numbers back as data, before you review the change.

[Set up your coding agent](https://streamlit.io/agents#skills)

Your next data app

- You Build a Streamlit dashboard over the orders table, with region and quarter filters.
- **On it, with Streamlit's skills.**
  Plain Python, cached queries, and filters you can share by URL.
- app.py metrics.py

Give your agent Streamlit's skills

`streamlit skills`

After installing the latest Streamlit.

Plain Python your team can read, review, and change.

For people and agents

## Write your business logic once. People get an app. Agents get a tool.

The app your team already uses in the browser doubles as a tool for Claude, ChatGPT, Cursor, or any MCP client. Agents go through the same logic and access rules as people do, and you don't write a line of extra code for it.

[How agents use your apps](https://streamlit.io/agents#tools)

- ### Ask the forecast app what happens at +5% price

  An assistant sets the price lever and reads the forecast back as data, computed by the app's own model.
- ### Let an agent draft the approvals for your review

  It fills in the review queue through the app's own forms and validation. You make the call.
- ### Find out why this week's numbers moved

  Ask in Claude or ChatGPT, and the answer comes from the dashboard's own numbers and definitions.

Made for data, built for dashboards

## Streamlit solves the hard parts of dashboards for every app

Hand a pandas, Polars, or Snowpark dataframe straight to a table or a chart. No serializing, no glue code. You get native charts for nearly every visualization, tables that load as you scroll, and cross-filtering and drill-down out of the box. We keep making them faster, and your apps pick that up when they upgrade.

advanced-dashboard-demo.streamlit.app/accounts [Open app ↗](https://advanced-dashboard-demo.streamlit.app/accounts)

The dashboard demo, built with Streamlit and running on Community Cloud. It may take a moment to wake up.

- ### Large data, fast

  Tables load rows as people scroll. Page sections load in parallel. Cached data refreshes in the background, so nobody waits on a stale query.

  - [`st.dataframe`](https://docs.streamlit.io/develop/api-reference/data/st.dataframe)
  - [`@st.fragment(parallel=True)`](https://docs.streamlit.io/develop/api-reference/execution-flow/st.fragment)
- ### Snappy interactions

  Only the section that changed reruns. Inputs can skip reruns until people apply their choices, so heavy queries run once.

  - [`@st.fragment`](https://docs.streamlit.io/develop/api-reference/execution-flow/st.fragment)
  - [`st.form`](https://docs.streamlit.io/develop/api-reference/execution-flow/st.form)
- ### Rich exploration

  Select points on a chart or rows in a table to cross-filter and drill down. Every filter and view can be shared by URL.

  - [`on_select="rerun"`](https://docs.streamlit.io/develop/tutorials/elements/dataframe-row-selections)
  - [`bind="query-params"`](https://docs.streamlit.io/develop/api-reference/caching-and-state/st.query_params)
- ### Native charts

  KPIs with sparklines, standard and statistical charts, maps, diagrams, and charts inside tables, all built in and styled by your theme.

  - [`st.altair_chart`](https://docs.streamlit.io/develop/api-reference/charts/st.altair_chart)
  - [`st.echarts_chart`](https://docs.streamlit.io/develop/api-reference/charts/st.echarts_chart)

[Explore dashboards](https://streamlit.io/dashboards)

In production

## From one team's dashboards to 500 production apps

- Open Insurance

  ### 30 dashboards moved in one month

  Two engineers moved 30 dashboards to Streamlit in one month, remodeling the data to consistent definitions on the way. The team now spends half as much time creating dashboards as in its previous custom environment.

  [Read the case study →](https://www.snowflake.com/en/customers/all-customers/case-study/open-insurance/)
- Snowflake

  ### More than 500 production apps

  Streamlit apps serve teams across Snowflake, from finance to security. The finance team plans ten years out in [Snowplan](https://www.snowflake.com/en/blog/fpa-ai-modernization-snowplan-v2/), a Streamlit app with editable assumptions, scenarios, and review and approval.

  [Read the write-up →](https://www.snowflake.com/en/engineering-blog/scaling-enterprise-analytics-streamlit-apps)

- ### Coming from a BI tool?

  Move the dashboards your teams rely on to Streamlit as a modernization, not a one-to-one copy. They gain your Python models and logic, writeback through that same logic, and access for AI agents. On Snowflake, they keep your governed data and access controls.

  [How modernization works →](https://streamlit.io/dashboards#modernize)
- ### Best on Snowflake

  Streamlit runs natively inside Snowflake. Your data stays on the platform, Snowflake's access controls apply, and Cortex Code builds the app for you.

  [Streamlit on Snowflake →](https://www.snowflake.com/en/product/features/streamlit-in-snowflake/)

## Frequently asked questions

- ### Is Streamlit production-ready?

  Yes. Streamlit apps run in production at companies of every size: inside Snowflake, in their own clouds, and on Community Cloud. Built in, you get authentication with any OpenID Connect provider ([`st.login`](https://docs.streamlit.io/develop/api-reference/user/st.login)), caching with background refresh ([`@st.cache_data`](https://docs.streamlit.io/develop/api-reference/caching-and-state/st.cache_data)), partial reruns ([`@st.fragment`](https://docs.streamlit.io/develop/api-reference/execution-flow/st.fragment)), theming, accessibility, headless testing ([`AppTest`](https://docs.streamlit.io/develop/api-reference/app-testing/st.testing.v1.apptest)), continuous security scanning, and a stable API with extremely rare breaking changes, so upgrades improve existing apps. [More on running Streamlit at scale](https://streamlit.io/enterprise).
- ### Should my coding agent build this data app in Streamlit or in React?

  For an internal data app, a dashboard, or an AI app over your data, use Streamlit. The agent writes plain Python instead of a frontend, a backend, and the contract between them. Streamlit is batteries-included, so charts, data tables, forms, authentication, caching, and theming are native features instead of a mix of separate libraries. That's less code to generate and maintain, fewer places to break, fewer tokens to pay for, and code your data team can review in the language it already knows. Every app is also a tool for other agents, at no extra cost. Choose React when the UI itself is the product, when the app is customer-facing at large scale, when it must work offline, or when it has specific design requirements that Streamlit's components and theming can't meet. [Compare them in detail](https://streamlit.io/why-streamlit#compare).
- ### Can Streamlit replace Power BI or Tableau?

  For the dashboards your teams rely on, yes: as a modernization, not a one-to-one copy. Streamlit ranked No. 1 among libraries for creating dashboards in the [Python Developers Survey 2024](https://lp.jetbrains.com/python-developers-survey-2024/) by the PSF and JetBrains. Together with Snowflake, it covers much of what teams use a BI tool for: governed metrics from semantic views, Snowflake's access controls, interactive filtering and drill-down, live refresh, and sharing and embedding.

  Moving a dashboard to Streamlit is also the chance to add what BI tools weren't built for: your existing Python models and forecasts, writeback validated by the same logic that produced the numbers, access for AI agents, and versioned code that agents build and maintain.

  Rebuild the dashboards that matter on governed metrics, check the numbers against the original, and retire the reports nobody opens. A BI tool still fits drag-and-drop self-service and paginated reports. [How modernization works](https://streamlit.io/dashboards#modernize).
- ### Can AI agents use my Streamlit apps?

  Yes. Every Streamlit app also serves agents, over MCP or plain HTTP, with no code changes. They read its tables and charts as data, set its filters, and submit its forms. Everything goes through the same logic and access rules as a person in the browser. Agents use the same caches, connections, and partial reruns as the app's human users, so data that's already cached isn't queried again. Add the app's URL to Claude, ChatGPT (through custom connectors), Cursor, or any MCP client. There's no second API to build, secure, and maintain, and one setting turns agent access off for an app. [How agent access works](https://docs.streamlit.io/develop/concepts/ai-agents/agent-access).
- ### Is Streamlit secure?

  Streamlit handles its own security centrally, for every app that uses it. Its code runs through static security analysis on every change and every week, and AI-driven security scans review the Python server and the web frontend regularly. Its own dependencies, including the entire web frontend it ships, are checked for known vulnerabilities daily. Fixes ship in regular releases, so upgrading Streamlit brings them to every app. Your app's own Python packages stay yours to manage, as in any Python project. Data access runs on the server, protection against cross-site request forgery is on, embedded HTML runs no scripts unless you opt in, and researchers report vulnerabilities through Snowflake's bug bounty program. On Snowflake, apps also run inside Snowflake's security perimeter and access controls. [More on security](https://streamlit.io/enterprise#security).

Get started

## Your next data app starts right here.

One install. Your coding agent, set up with Streamlit's skills. Then ask for the Streamlit app you need, or write it yourself.

[Follow the quickstart](https://docs.streamlit.io/get-started)

Ready to share? Deploy public apps for free on [Community Cloud](https://share.streamlit.io/?utm_source=streamlit&utm_medium=referral&utm_campaign=main&utm_content=-ss-streamlit-io-getstarted), or run private apps [in Snowflake](https://signup.snowflake.com/?utm_source=streamlit&utm_medium=referral&utm_campaign=main&utm_content=-ss-streamlit-io-getstarted).

Your terminal

1. 01Install Streamlit.

   `pip install streamlit`
2. 02Give your coding agent the skills.

   `streamlit skills`
3. 03Or say hello to your first app.

   `streamlit hello`

You're set. Ask your agent for the Streamlit app you need.

Just looking around? [Try Streamlit in your browser ↗](https://streamlit.io/playground)
