AI Chat for the database work already open in Streams

Generate SQL, repair a failed query, or ask why a migration stream is unhealthy — without leaving DBConvert Streams or copying context into a separate chat.

Get DBConvert Streams AI Chat setup

See the tools behind each answer while the chat works.

Generate SQL

Describe the result you need in plain English. AI drafts the query in the SQL editor against the database you selected.

See how

Fix failed SQL

The failed statement and database error are already attached. AI can inspect the relevant structure before proposing a correction.

See how

Talk to your live data

Ask what is in a table, why a CDC stream is failing, or whether an export still matches the source. Chat starts with the item already open.

See how

Context is already there

Open AI Chat from a table, stream, file, or SQL editor. It starts scoped to the thing you are working on.

No copy-pasting

It can inspect the connected schema, current stream status, query error, or file metadata before it answers.

Use the model account you already have

Choose from supported agents such as Claude Code, Codex, Copilot CLI, or OpenCode. Streams shows tool activity and cost while it works.

What a standalone AI chat can’t see

For a one-off query on a schema you can paste into ChatGPT, Claude, or Copilot, they are useful. The difference starts when the answer must reflect the database, data, or CDC stream in front of you right now.

A standalone AI chat compared with DBConvert Streams AI Chat for live database work.
What changes	ChatGPT, Claude, or Copilot — on their own	DBConvert Streams
Uses the current schema and data	Works from the schema, rows, and constraints you paste or attach.	Reads the connected schema and can run bounded read-only queries against the live workspace.
Fixes a failed query with evidence	Needs you to copy the SQL, database error, and relevant context into the chat.	Starts with the failed SQL and exact database error, then checks the relevant schema before suggesting a fix.
Diagnoses a CDC stream	Cannot see the stream, its run history, or its logs unless you provide them.	Reads stream status, statistics, recent errors, and recent log entries from the workspace.
Keeps context current	Pasted context can be incomplete or stale as soon as the database changes.	Each tool call reads the current workspace state instead of a second copy of your schema.
Protects production data	Safety depends on the prompt, connected tools, and permissions you give it.	The server accepts only SELECT and WITH … SELECT queries; it cannot change data, connections, or stream state.

Need AI in the app or in your own client?

Both read the same workspace. Choose AI Chat when you are working in Streams; choose MCP when the work happens in your IDE or AI client.

	AI Chat	MCP for AI clients
Where you work	Inside the DBConvert Streams desktop app	In your own AI app or IDE
What it needs	A supported agent CLI installed and signed in	Any MCP-compatible client, plus a click or an API key
Deployment	Desktop app only	Desktop (stdio), Docker or remote server (HTTP)
Toolset	Scoped to what you opened the chat on, so it stays focused	The full workspace toolset
Editor integration	Generate, fix and insert SQL in the Streams editor	Whatever your client offers
Who pays for the model	Your agent account — cost shown in the chat	Your client’s own account

Explore MCP for AI clients

Answers that cannot alter production

AI Chat can advise, generate and place SQL in the editor. Database inspection, queries and stream diagnostics remain read-only; it cannot edit connections, start streams or modify data.

Set up AI Chat

Want to use Claude, Codex or Cursor outside Streams? Use the MCP server instead.