Oxys DB Studio

Open source · pre-release

The MongoDB workbench that stays out of your way.

Oxys DB Studio is a native desktop IDE for MongoDB, written in Rust. Browse, query, aggregate, import, compare and automate, and script every one of those actions from the command line.

Download coming soon Explore the features

Linux and Windows builds in testing · macOS planned · MIT or Apache-2.0

Oxys DB Studio showing the products collection of a sample store database in a table, with the connection tree on the left and a query bar above the results.
The same view in the dark theme: the products collection of a sample store database in a table.

Why Oxys DB Studio

Built for people who keep a database client open all day.

  • Native and quiet

    Compiled Rust with a GPU-drawn interface: nothing to install besides the app, no Java or JavaScript runtime. It draws a frame only when something changes; with a collection open and nothing happening, our Linux release build measured 0.2–0.6 % CPU.

    How it is measured

  • Careful with your data

    Drops, bulk deletes, overwrites and syncs ask first. A read-only connection is enforced by the engine itself, not just by greyed-out buttons. Malformed documents and dropped connections produce errors you can act on, not crashes.

  • Every action is a command

    Each menu item, button and shortcut runs a named command with JSON parameters. The command-line client and the local control channel run the same commands. A local MCP server, off by default, gives AI agents read-only tools.

  • Open source

    The application, the command-line client, the team server and the task runner are published under the MIT or Apache-2.0 licence, at your choice. Read the code, build it, change it.

  • Team features you host

    Share connections, queries and scripts through a team server you run yourself, and run scheduled tasks on your own task runner. Docker Compose files and a Helm chart are included.

  • Private by design

    No telemetry, no analytics, no crash reporting and no account. The application only connects to the databases you choose, plus an update check you can turn off. Every outbound connection is listed.

    What it connects to

A quick tour

From the first query to the nightly export.

Query and edit documents

Type a filter, projection and sort, or build them visually by dragging fields into the query builder. Results page through large collections in a table, tree or JSON view, and you edit values in place.

Querying in detail
The visual query builder next to a collection's results, with a condition on the category field.

Build pipelines stage by stage

The aggregation editor shows each stage on its own, with its input and output, so you see where a pipeline goes wrong. Explain it, export its results, or turn it into code.

Aggregation in detail
The aggregation editor with match, group and sort stages on an orders collection, and the pipeline output below.

Ask in SQL, get MongoDB

Write SELECT … WHERE … GROUP BY against your collections. The query runs on the server as a MongoDB query, and the Query Code tab shows exactly which one.

SQL queries in detail
A SQL query on the orders collection and its results in a table.

A shell when you need one

IntelliShell runs the bundled mongosh with completion and syntax highlighting, and shows results in the same table, tree and JSON views as everything else.

The shell in detail
IntelliShell with a short script and its result.

See what is really in a collection

Schema analysis samples documents and reports each field's types and how often it appears, with charts. Explain plans and the index manager help you make the slow queries fast.

Schema and performance
Schema analysis of an orders collection: the list of fields with their probability and type, and a chart of type distribution.

Scriptable

What you can click, a script can run.

The desktop app is one client of an engine of named commands. The command-line client is another: list the commands, run one with JSON parameters, or pipe a whole session through it, one command per line.

AI agents do not get every command. The local MCP server, off until you turn it on, gives them nine read-only tools (list connections, databases, collections and indexes; find documents; explain; analyse a schema; collection statistics; list connected tabs) on 127.0.0.1, with a token, for the saved connections you allow.

AI assistant and MCP 
# every command, with its parameters
$ oxysdb-cli commands

# run one
$ oxysdb-cli run connection.new \
    '{"uri": "mongodb://db.example.internal", "name": "Staging"}'

# one session, one command per line
$ oxysdb-cli batch < session.jsonl

# read-only tools for local AI agents
$ oxysdb-cli mcp --allow-connection Staging

Self-hosted

Your team server, on your infrastructure.

Team Sharing keeps shared folders of connections, queries and scripts, with manage, edit and view permissions, on a server you run. Passwords are not shared unless you choose to. The task runner, a separate service, runs saved exports, imports and migrations on a schedule, next to your databases.

The desktop app, the team server and the runner talk over TLS 1.3 with a hybrid post-quantum key exchange (X25519MLKEM768) and 256-bit ciphers, and you can require it on both ends. What is post-quantum, and what is not yet.

Start with one container and a SQLite file. Grow to several replicas on MongoDB or on FerretDB with PostgreSQL.

Self-hosting guide
Desktop appyou and your team Desktop appa colleague Team servershared folders, permissions Task runnerscheduled tasks Your databaseson your network
Desktop apps share through a team server and send tasks to a runner. Both services run on your infrastructure; the desktop app also works fully offline without them.

Platforms

Where it runs

No release has been published yet. These are the packages the build produces and tests today.

  • Linux

    x86_64 and arm64: .deb, .rpm and a tarball. Needs glibc 2.35 or later.

    In testing

  • Windows

    x86_64: installer, MSI and zip. Not code-signed yet. An arm64 build is experimental.

    In testing

  • macOS

    Apple silicon and Intel builds are planned, signed and notarised.

    Planned

  • Servers

    Team server and task runner as container images, with Docker Compose files and a Helm chart.

    In testing

Follow the road to the first release.

Read the code, try a build from source, or open an issue with what your team needs.

Source code Documentation