Dig into your databases. DGopher is a fast, free and open-source database client for Linux, macOS and Windows. One app for PostgreSQL, MySQL, SQL Server, Oracle, ClickHouse, SQLite, DuckDB and Redis, and for CockroachDB, YugabyteDB, TimescaleDB, MariaDB and TiDB as engines of their own.
Download · Features · Engine support · Documentation
- Native and light. One Go binary, drawn by MyGo, with no webview, JVM or Electron. It starts instantly and looks the same on every system. See Footprint.
- Hard to break production. Mark a connection as production and its color shows everywhere. Writes ask first, a destructive statement needs the connection's name typed, transactions commit only when you say, and an open transaction is never closed silently. See the Safety model.
- AI that works with your query. Ask Claude Code, Codex or pi beside
any editor. It runs the
claude,codexorpicommand line you have installed, so it uses your existing subscription and its models, with no API key to give DGopher. It sees your statement and its last error, reads the schema, and writes SQL you insert with a click. It runs a query on its own only after you allow it, only reads, never on production, and has no access to your files or shell. - SQL, Redis and files in one app. Query SQL databases, browse and edit Redis on one server, a cluster or through Sentinel, and drop a CSV, Parquet or JSON file on the window to query it with DuckDB.
- Made for teams. Connections, query files, dashboards, data models and snippets live in your Git repository, without passwords. Every statement and edit goes into a tamper-evident audit log.
- Each engine as itself. Plans, catalog, server activity and backup
follow each engine's own tools, from PostgreSQL's
EXPLAIN ANALYZEto Oracle's PL/SQL and Data Pump, SQL Server's T-SQL batches and TimescaleDB's hypertables. See Engine support.
DGopher is young and has not reached 1.0. Known limitations lists what it does not do yet, and the Roadmap what is planned.
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
Download the build for your system from the latest release:
| System | Builds |
|---|---|
| Linux | amd64, arm64 (.deb, .rpm and .tar.gz) |
| macOS | Apple silicon, Intel (disk image) |
| Windows | amd64 (installer) |
The builds are not signed for macOS or Windows yet. On macOS, allow the app once in System Settings → Privacy & Security. On Windows, SmartScreen warns once.
The app checks for updates once a day and offers to install them, or
choose Check for Updates… in the menu. The .deb and .rpm are
updated by installing the new one instead.
To build from source, see Development.
- Open DGopher and choose Try the sample on the start page. It creates a small shop database in SQLite, so you need no server.
- Take the one-minute tour the start page offers.
- Press ⇧⌘P (Ctrl+Shift+P on Linux and Windows) for the command palette, or ⌘N
for a new connection. You can paste a connection URL such as
postgres://user@host/dbinto the form, or choose Neon or Supabase.
- Editor: completion from the catalog, a ▶ on every statement,
problems underlined as you type, hover cards that say what a name is, Go
to Definition, Find Usages and Rename,
:nameparameters, multiple carets, Vim keys, and a session per editor. - Results grid: streams large results, chooses blocks of cells as a spreadsheet does, filters in SQL or with a builder, sorts, follows foreign keys both ways, compares two results row by row, profiles every column, and opens values in JSON, XML, hex and image viewers.
- Edit rows safely: edits in the grid are shown as SQL to review, then applied in one transaction, one row per statement by primary key.
- Query plans: Explain and Explain Analyze as a tree of steps or a flame graph, with rows estimated and found, and advice on what makes a query slow.
- Charts and dashboards: chart any result, and put any read query on a dashboard as a number, a chart or a table, with parameters and refresh.
- Notebooks: show any SQL file as cells, Jupyter-style, each with its own results, with Markdown between them. The file stays SQL that runs as a script.
- Files: import CSV, JSON, Parquet, Excel and XML. Export to CSV, JSON, SQL, Markdown, Excel, Parquet and more.
- Move data: copy tables between databases and engines, compare rows, generate test data, back up and restore, and run SQL dumps of any size, checked before they run.
- Schema: ER diagrams you can edit, a table designer, data models that generate migrations, a query builder, search across objects and their definitions, and documentation of a schema as HTML or Markdown.
- Admin: server activity with running queries to cancel, locks and the slowest statements; users and privileges; and VACUUM, ANALYZE, REINDEX or OPTIMIZE from a table's menu.
- Redis: a key browser, value editors for every type including Redis 8's, a console with completion, and monitor, slow log, memory and Pub/Sub panels.
- Passwords stay in the system keychain or a
DGOPHER_*environment variable, come from a command such as a password manager, or are asked every time. Secrets are redacted from history and logs. - Log in without a password through AWS IAM, Google Cloud IAM or Microsoft Entra ID, or Kerberos on PostgreSQL and SQL Server.
- TLS with a custom CA and client certificates, SSH tunnels through jump hosts with host keys checked, and SOCKS5 or HTTP proxies.
- Read-only connections are enforced in the app and on the server (on TiDB, which refuses the server's setting, by the app alone).
- Columns such as
password,tokenorcard_numbershow••••••, and a connection can keep its rows from the AI chat.
- Keyboard first: a command palette, quick open for tables and query files, and keys like Zed's, every one of which can be changed.
- History and snippets: every statement run, with its timing and error, a key away; snippets with fields filled in by Tab.
- Tabs and groups: pin, reorder and reopen tabs, split them side by side, drag one into another window, and get them back after a restart.
- Errors that help: an error says what most likely went wrong and offers the fix, such as Retry or Edit Connection.
- Your way: light, dark and custom themes, zoom, and every control named for screen readers.
Features describes everything DGopher does.
Measured on Linux (amd64, Wayland) with version 0.7.0, at the start page:
| Download | 45 MB, as the .deb or the .tar.gz |
| Executable | 131 MB, with every driver and DuckDB compiled in |
| Memory at start | about 180 MB resident, of which about 55 MB is the app's own; the rest is the executable and the GTK and graphics libraries, mapped from disk |
The documentation index lists every page:
- Features, Engine support and Keyboard shortcuts
- Safety model and Audit log
- Where things are kept and Project files
- Command line and Troubleshooting
- Known limitations and Roadmap
- For contributors: Development, Architecture, Testing, How SQL is read and CONTRIBUTING.md
Most of DGopher's code and documentation is written with AI coding agents, such as Claude Code and Codex. The maintainer directs, reviews and tests every change.
See CONTRIBUTING.md. Report vulnerabilities privately, as SECURITY.md describes. Each version's changes are in its release notes.
Copyright 2026 DGopher contributors. Licensed under the Apache License, Version 2.0.
The icon is based on the Go gopher, designed by Renée French and licensed under CC BY 4.0.










