Let your coding agent install AddisDB, then use it safely

How it works

Paste one line into your agent. Claude Code, Codex, Cursor or VS Code. The agent reads the setup page, tells you what it will install, and asks before it does.

It installs and connects AddisDB. One command installs the CLI, the desktop app and the editor extension, verifies every download against a published checksum, and registers AddisDB’s MCP server with your agent. No admin rights.

Your agent works with your databases, safely. It lists your connections, reads schemas, runs read-only queries, watches live activity and previews migrations, without ever seeing a password. Ask it to show you, and AddisDB opens the query, table or ER diagram while you watch, in the desktop app, VS Code or the terminal.

Safe by default

Read-only until you approve: Agents read through the same safety layer the app uses. Any change, from a one-row update to a restore, runs only after you approve it in AddisDB, with its SQL and risks in front of you.

Always visible, one click to stop: While an agent is connected, AddisDB shows a banner with its name, its activity, and a Stop button. Stop abandons whatever it was running and cuts it off.

Your rules: One settings page decides what every agent may do: switch agents off entirely, approve reads too (always, or only on prod), or refuse changes outright. It applies from the next thing an agent tries.

No credentials: Connections are referenced by name. No tool result contains a password, token or connection string.

Capped results: 200 rows by default and long values clipped, so a stray SELECT * doesn’t pour a customer table into a model.

Sensitive columns masked: Mark a column sensitive in AI Privacy and its values come back masked to every AI, coding agents included. Tools that can’t mask refuse instead. You choose the columns; nothing is guessed.

Migration preview and a sandbox: Agents are told to preview every schema change and show you the risks, and can try it on a disposable copy of the database before it touches the real one.

Free for local databases: Agent features work on every plan. Databases on your own machine are free; remote ones follow the usual plans.

What your agent can do

• addisdb_list_connections: Saved connections with engine and environment tag. Never credentials.
• addisdb_schema: Tables, columns, keys and row estimates.
• addisdb_query: Read-only SQL, capped rows, timing.
• addisdb_explain: The query plan and cost, without running it.
• addisdb_preview_migration: Destructive statements and warnings, with severities. Nothing executes.
• addisdb_inspect: Definitions, exact counts, cost, drop impact, the server’s databases and users.
• addisdb_monitor: Live sessions, server metrics and the statements costing the most time.
• addisdb_compare: One query across several databases, diffed, to catch drift.
• addisdb_sandbox: A disposable copy of a database to test a migration on first.
• addisdb_open: AddisDB on any view, in the desktop app, VS Code or the terminal.
• addisdb_show_query: Your agent’s SQL in an AddisDB tab, run while you watch (read-only).
• addisdb_show_table: A table’s data in the grid, optionally filtered.
• addisdb_show_erd: The ER diagram, focused on the tables it’s discussing.
• addisdb_get_view: What you have open, so “fix this query” just works.
• addisdb_apply_migration: Any change to data or schema: you see the SQL and its risks in AddisDB and click Approve.
• addisdb_drop_table: A drop that keeps a safety copy, with a one-call undo. You approve both.
• addisdb_import_csv: CSV import, plus export, backup and restore, each approved by you.
• addisdb_generate_mock_data: Realistic test rows, tracked so they can be cleared again.