AI for Database is positioned as being for teams without a data team, letting them ask data questions in natural language and get answers immediately from a live database or spreadsheet. Users can ask questions in plain English (natural language), without writing any SQL, and get instant answers, and turn any answer into a shareable dashboard or scheduled report with one click, and trigger alerts when the data changes. The official emphasis is no credit card required, try it without connecting, and read-only by default.
Main capabilities
- Natural language database querying: after connecting PostgreSQL, MySQL, MongoDB, or Google Sheets, ask questions directly in natural language, and the system automatically writes SQL and runs it on the database, returning the answer, the raw rows, and the query used. It supports follow-up questions and export to Sheets or CSV.
- Dashboards and reports: any answer can be turned into a shareable dashboard, charts are generated automatically, and scheduled reports and one-click share links are supported.
- Database-driven workflows and alerts: get notified when data changes, with support for email notifications and Webhook triggers. Workflows are described in a way similar to explaining them to a colleague, with no node graph required.
- Use through AI assistants: you can connect an AI for Database account in Claude, ChatGPT, Grok, and Muse, and then directly ask questions in natural language, run read-only SQL, and set alerts in chat. The official documentation gives connection steps for each platform (connectors in the Claude directory or settings; for ChatGPT, enable developer mode and create an MCP app at chatgpt.com/plugins and paste the server URL; for Grok, create a new custom connector at grok.com/connectors; for Muse, search in connectors). After connecting, you can: ask questions in natural language, understand an unfamiliar database (read the schema and explain tables, columns, and relationships), safely run your own SQL (run in a read-only transaction), investigate metric changes, set data alerts (such as a weekly Monday email reminder for tickets that have been open for more than 7 days), and view dashboards and remaining AI credits. Queries are read-only; workflows can send email or Webhook alerts, and approvals can be set in the assistant.
- Public REST API: use scoped API keys and an OpenAPI contract to inspect the schema, return raw query results, build paused workflows, preview recipients, test a single delivery, and explicitly publish. The process is divided into 6 steps: Discover (lists sanitized connections, health status, and cached schemas), Validate (runs read-only SQL and checks raw columns, rows, counts, and elapsed time), Draft (creates a paused workflow and references write-only outbound credentials), Preview (executes draft queries without touching external systems), Test (explicitly confirms one live operation and checks its sanitized attempt record), Publish (activates an immutable version and checks scheduled run history).
- Trust and control: choose self-hosting (all data stays on your own infrastructure) or cloud hosting (encrypted in transit and at rest). Permissions are controllable, starting from read-only, with specific operations enabled as needed. Rules can be set so high-risk operations require approval, while safe workflows can run automatically. Full self-hosted deployment is supported, with a one-time installation fee plus an annual license. Data at rest is encrypted with AES-256, is not stored in plaintext, and keys can be rotated at any time. Every query, dashboard, and operation is logged, and audit logs can be exported for compliance.
- Pricing and credits: simple plans are offered, and credits belong to the user. Pro is billed per user per month (settled monthly) and includes querying databases in natural language, exploring answers with charts and dashboards, saving queries and inspecting SQL, automating with scheduled workflows, triggering operations via webhooks, and inviting colleagues to the workspace. The Team plan includes everything in Pro, plus 2.5x per-user credit funding, separate credits for each member, assigning connections to team members, tracking usage by member, reviewing activity via audit logs, and sharing dashboards and workflows. The Enterprise plan is customized around business needs, with discussions available for custom usage and seat requirements, company SSO requirement review, security and compliance discussions, deployment planning with your team, tailored support terms, and procurement and contract support. Each user has a separate credit balance, and when paying monthly or annually, the full payment is converted into credits. Credits never expire, there are no weekly limits or monthly resets, and unused credits remain in the individual balance. You can top up $10, $25, $50, or $100 at any time, and AI usage pauses when the balance runs out.
How to use Connect a database or spreadsheet, ask questions in natural language, pin answers as dashboards, and then set up automated workflows and alerts. You can also connect through AI assistants (Claude, ChatGPT, Grok, Muse) and use it directly in chat. You can also use the public REST API to discover schemas, validate queries, draft and publish workflows programmatically.
Use cases Officially listed: operations teams ask "which transactions are stuck for morethan 14 days?", generate a table + chart + weekly report, and notify the owner on Slack; product managers ask "which SKUs are low in stock?", generate an inventory dashboard, and email the operations team; founders/general managers ask "how long is our cash runway?", generate a financial health dashboard and a weekly executive report; product managers ask "which features drive retention?", perform usage analysis and trends, and sync insights to the product roadmap.
Limitations and notes Queries are read-only by default, workflows can send email or webhook alerts, and approvals can be set in the assistant. The official emphasis is no credit card required, a free model that remains available permanently, and read-only from the first handshake. Details such as the specific free quota, exact prices for each plan, credit consumption rules, and supported database versions are not all provided in the materials and should be based on the official page.
