AI That Reads Your Live Databases

Open AI Chat Connect an External Client

Talk to your database

Ask questions about your data. Get answers from your live workspace.

Generate SQL

Describe what you need in plain language. AI writes the SQL.

Fix failed SQL

AI reads the error and suggests a corrected query.

AI Chat — DBConvert Streams

Ask a question, watch the read-only tools it runs, then read the answer — all inside the app.

Two Features

Which One Do You Want?

They are not two doors to the same room. AI Chat is a feature of the app. MCP is a server the app exposes to everything else.

AI Chat

An assistant that lives in the DBConvert Streams desktop app and knows what you're looking at. Open it from a table, a file, a stream config, or a SQL console and it starts already scoped to that thing.

Runs the agent CLI you already have — Claude Code, Codex, Copilot CLI, or OpenCode, on your own account
Nothing to register: the tools are injected into the session for you
Writes SQL straight into the editor with ⌘K, and repairs a failed query with the error already attached
Shows every tool call as it runs, plus token use and cost per chat

Set up AI Chat

AI Clients (MCP)

A read-only MCP server that hands your workspace to the AI tool you already work in — so the assistant writing your migration, model, or query is reading the real database, not guessing from your ORM file.

One-click Connect for 8 clients from the desktop app, or an HTTP endpoint for Docker and remote servers
The full toolset, unscoped — your client can roam the whole workspace
Your database context sits next to your code, repo, and files in the same session
The only path that works when DBConvert Streams runs on a server, not your desktop

Connect a client

	AI Chat	AI Clients (MCP)
Where you work	Inside the DBConvert Streams desktop app	In your own AI app or IDE
What it needs	At least one supported agent CLI installed and signed in — add more and switch between them anytime	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	All 27 tools, every time
Editor integration	⌘K generation, Fix with AI, Insert into editor, Keep / Revert	Whatever your client offers
Who pays for the model	Your agent account — cost shown live in the header badge	Your client’s own account

Nothing stops you running both. They read the same workspace, so an edit in the UI shows up in the next answer on either side — no restart, no second copy of your schema.

Use Cases

What People Actually Ask It

You describe the goal; the AI picks the tools, chains them, and comes back with the answer and the evidence.

The overnight stream failure

Both

Instead of opening the run history, scrolling logs, and cross-checking the config, ask once. It pulls status, stats, recent errors, and the stream config together and names the cause.

“My MySQL → PostgreSQL CDC stream stopped overnight. Check status and recent errors, and tell me what to fix.”

Catch a migration blocker first

Both

Before you start a 10M-row load, have it diff the two schemas across connections. It flags type mismatches, missing primary keys, and the things that quietly break CDC.

“Compare orders on my Postgres source with the MySQL target. What will break, and what is only a warning?”

Prove the data actually landed

Both

One federated SELECT spans a database, local files, and S3 at once — so you reconcile an export against the live table without a staging table or a throwaway script.

“Is customers.parquet in S3 still in sync with the live customers table? Row counts, and rows on only one side.”

Give your IDE assistant the real schema

MCP

Writing a migration, an ORM model, or a report query in Cursor or VS Code? The assistant reads the actual columns, types, and indexes from the live database instead of inferring them from stale code.

“Write the migration to add this column, matching the types already in the production orders table.”

All of it runs on 27 read-only tools — schema, SQL, federated queries, files & S3, streams, comparison. Plain-language map in the tools reference .

Live Sessions

Real Prompts, Real Answers

Every exchange below came back from a live workspace — numbers and all, on the inexpensive Claude Haiku 4.5.

Stream troubleshooting

You

My MySQL → PostgreSQL CDC stream stopped overnight. List my streams, check status and recent errors, and tell me what to fix.

Found it — the latest run of 10M-cdc-mysql2pg FAILED with Error 1049 (42000): Unknown database 'target', flagged source-side. The CDC source points at a database that doesn't exist on the MySQL server. Fix: create it (or correct the database name on the connection), then restart.

Analytics on live data

You

What are my top film categories by rental revenue? (sakila sample)

Sports — $5,314.21 Sci-Fi — $4,756.98 Animation — $4,656.30 Drama — $4,587.39 Comedy — $4,383.58 Joined category → film_category → film → inventory → rental → payment, ran in ~55 ms. Want the query to reuse, or the full top 15?

Trimmed for the page — full transcripts: plain-English SQL and federated validation.

Safety

Read-Only by Design

The same boundary on both paths. The AI can look and advise. It cannot change anything.

SELECT-Only Filter

Only SELECT and WITH … SELECT pass the server-side safety filter. INSERT, UPDATE, DELETE, DROP, ALTER, and CREATE are rejected at the server before they reach your database.

No Credential Access

Connections are read from your workspace; the AI never sees passwords and cannot create or edit connections. Secrets are redacted from everything it reads.

No Stream Control

The AI can't start, stop, pause, or reset streams. It surfaces the status, the error, and the fix — you (or the UI) run the action.

Your Infrastructure Only

The server runs on your machine or in your stack and only answers your client's requests. It never calls out to DBConvert or your AI provider on its own.

A fast analyst on tap — not an autopilot for your production database.

Full boundary in Safety & privacy.

Setup

Getting Started With Either One

Two features, two different setups. Neither takes longer than a few minutes.

AI Chat

Desktop app

There is no MCP registration on this path — DBConvert runs an agent you already have and injects the tools for you.

1Install one agent CLI and sign in with your own account.
2Fully quit and reopen DBConvert Streams so it finds the new command.
3Click AI Chat in the sidebar — or Ask AI on any table, file, stream, or SQL console to start already scoped to it.