ChatGPTWren AI for ChatGPT
A ChatGPT data agent that knows your business.
Add Wren AI to ChatGPT and ask in plain language. Every answer comes from live data, with the SQL, a chart, and the definitions behind it.
- Live data, no exports
- SQL on every answer
- Row- and column-level security
ChatGPT, Wren AI appAnswered by Wren AI
Two regions missed target in Q3. APAC closed at 91% of target and LATAM at 96%. North America and EMEA beat it.
SELECT region,
SUM(net_revenue) / SUM(target) AS attainment
FROM sales_performance
WHERE fiscal_quarter = '2026-Q3'
GROUP BY regionIllustrative answer on a sample dataset.
Trusted by data teams worldwide, with 17,841 GitHub stars
Why it matters
A data agent is only as good as the definitions behind it.
A model with a database connection can write SQL. Whether that SQL means what your business means is a separate problem. Wren AI puts a context layer between ChatGPT and your data, so every answer resolves against definitions your team agreed on.
| Dimension | ChatGPT on a raw schema | ChatGPT with Wren AI |
|---|---|---|
| Business definitions | Inferred from table and column names | Resolved from your context layer, versioned in git |
| Joins | Chosen by the model on each question | Declared once and reused by every caller |
| Sources | One connection per database | 20+ sources through one app, with no ETL |
| Access | Whatever the connection's credentials can see | Row- and column-level security per user, at query time |
| Ambiguous questions | Picks one reading and commits to it | Asks which metric you meant when two could fit |
| Every answer | A number in a chat | The SQL that ran, a chart, and the context it used |
The same holds for any data agent in ChatGPT: answers are only as consistent as the business definitions it can read. Wren AI keeps those definitions in one place for every assistant your team uses.
How it works
From the app directory to a governed answer.
No config file to edit and no warehouse password pasted into a chat. Database credentials stay in Wren AI.
1. Model once
Define what your data means
Metrics, relationships, and business rules live in MDL, Wren AI's context model, as git-versioned files in your own repository. Change a definition in a pull request and ChatGPT uses it on the next question.
github.com/acme/wren-contextknowledge/metrics/net_revenue.yml
mainfinance-data: exclude refunds from net revenue (#42)1name: net_revenue2label: Net revenue3description: Revenue after refunds, before tax4expression: SUM(orders.amount) - SUM(orders.refund_amount)Merged in a pull request. ChatGPT uses this definition on the next question.5owners: [finance-data]2. Connect
Add Wren AI to ChatGPT
Open Wren AI in the ChatGPT app directory, add it, and connect it to your Wren AI project. Access is scoped to that project, and you can revoke it without touching your database.
chatgpt.com/appsChatGPT appsWren AI
App for ChatGPT
Ask questions about your business data and get governed answers, with the SQL and a chart.
ConnectDataBusiness- Answers from live data across 20+ sources
- Uses your team's definitions, not the raw schema
Connected to project: sales-analytics3. Ask
Get answers you can check
Ask in your own words. Wren AI queries live data and returns the SQL that ran, a chart where it helps, and the definitions it used. When a question could mean two metrics, it asks which one you meant.
ChatGPTWren AIHow did net revenue trend by month this year?
Used Wren AI
Net revenue grew in 8 of 9 months, from $1.1M in January to $1.6M in September.
SELECT DATE_TRUNC('month', order_date) AS month, SUM(net_revenue) AS net_revenue FROM orders GROUP BY 1 ORDER BY 1metric: net_revenuerule: fiscal calendarAnd bookings?
Used Wren AI
Two metrics could fit. Which one did you mean?
Gross bookingsNet bookings
One context layer
Every assistant your team uses, one definition of revenue.
Teams rarely standardize on a single assistant. When each tool connects to the database on its own, each carries its own idea of what the numbers mean. Wren AI defines them once and serves them to every surface, with the same access rules.
ChatGPT app directory
Ask in the chat your team already works in, and get the SQL and a chart back.
Claude connector, MCP
Turn one prompt into a dashboard where every number links to its SQL.
Claude dashboards →SlackSlack app
Answer the question in the thread where it was asked.
Wren AI in Slack →Microsoft TeamsTeams connector
Bring governed answers into Teams chats and channels.
Wren AI in Teams →Your own agentAPI, MCP server
Give your product or internal agent the same context layer.
MCP server →FAQ
Wren AI in ChatGPT, answered.
Open Wren AI in the ChatGPT app directory, add it to ChatGPT, and connect it to your Wren AI project. If you don't have a project yet, create one in Wren AI and connect a data source first. Then ask your first question in plain language.
A data agent that reads a warehouse directly has to infer what tables and columns mean. Wren AI adds the layer that settles it: metric definitions, declared joins, and business rules, written once in a git-versioned context model. The same definitions then serve ChatGPT, Claude, Slack, Teams, and your own agents, so the answers agree. Wren AI also reaches 20+ sources, including operational databases such as PostgreSQL, SQL Server, Oracle, and MySQL.
Live data. Wren AI queries your connected sources when you ask, so the answer reflects what is in the database now. There is no ETL step and no data migration.
Row-level and column-level security are enforced in Wren Engine at query time, based on who is asking. The rules you already set apply no matter which chat window the question came from. Read how Wren AI enforces role-based access control on AI-generated SQL.
20+ sources, including PostgreSQL, SQL Server, Oracle, MySQL, Snowflake, BigQuery, Databricks, Redshift, ClickHouse, Athena, Trino, Starburst. One Wren AI project can span several of them, and the engine plans each query against the source where the data already lives.
Yes. Wren AI core is open source on GitHub, so you can inspect the engine and the context model and run them against a sample database first.
Try it
Give ChatGPT your business definitions.
Add Wren AI from the ChatGPT app directory, or see it on your own data with our team.