FluxDB

A native, cross-platform database client and SQL editor designed for high performance and low overhead: query, edit data, manage schemas and debug Redis in one window.

FluxDB is built with Rust and GPUI, without Electron or a WebView. It brings MySQL, MariaDB, PostgreSQL, KingbaseES, SQL Server, SQLite, TiDB, Dameng DM8, MongoDB and Redis into one desktop workflow, with a permission-aware AI Agent.

English | 简体中文

Download the latest release · Latest changes · Features · Database support · User guide · Report an issue

Latest changes

v0.1.18

• Server panels: a connection's context menu opens a "Server" tab with overview, sessions and lock waits for MySQL / MariaDB / TiDB, PostgreSQL / KingbaseES, SQL Server, Oracle and Dameng; Redis adds clients and slow log, and MongoDB adds current operations and an in-session trend. Queries, sessions and clients can be terminated after confirmation, read-only connections are blocked at three layers, and panels refresh manually or every 5 seconds.
• Switch query scope: the toolbar switches the target database and schema per connection type without opening a new query. Database switches clear results, running queries block switching, and PostgreSQL / KingbaseES return to the default schema with RESET search_path.
• Custom themes and fonts: create, edit and delete themes with SQL highlight tokens, import Zed / VS Code themes (JSONC included), and set UI font/size and code font/size/line-height independently with live preview.
• Local MCP over in-app HTTP: the stdio helper process is replaced by a loopback-only Streamable HTTP endpoint with a Bearer token, and existing client configs are taken over automatically.
• Improvements and fixes: centralized SQLite migrations with VACUUM INTO backups and fail-closed version checks, unified SQLite / SQL Server / TiDB value and read-only semantics, numeric Redis database sorting, and exact negative DECIMAL rendering on SQL Server.

Full release notes · All releases

v0.1.17

• Structure compare and data transfer: new connection context-menu entries for structure compare/sync and cross-connection data transfer. Tables are paired by schema.name and reported as missing, extra, modified or identical with side-by-side DDL, then the selected tables can be created, dropped or recreated. Transfer supports append and truncate-then-write, creates a missing target table, and reads/writes in pages with cancellation.
• Task center: collects backup, restore, SQL file execution, table and MongoDB import/export, structure sync and data transfer, with unified status, progress and logs, cancellation, and persisted history.
• Visual query plans: the toolbar "Explain" action parses MySQL / MariaDB / TiDB, PostgreSQL / KingbaseES and SQLite plan output into a node tree and inspector, flagging full table scans, filesorts and missing indexes, and falls back to the raw result when parsing fails.
• Connection form improvements: high-frequency fields stay under "Connection info", MySQL-family charset/collation/URL parameters move to "Advanced", and SQLite hides the empty network tab. MySQL / MariaDB / TiDB gain SOCKS5 and HTTP CONNECT proxies combinable with TLS, and PostgreSQL reuses the shared proxy dialer.
• Fixes: corrected legacy MySQL TLS-mode migration, kept compatible filter values when switching fields, added missing backup/restore cancel strings, and fixed KingbaseES completion compatibility and a nested block_on in export pagination.

Full release notes · All releases

Why FluxDB

• One workspace for multiple data sources: manage relational databases, MongoDB and Redis side by side.
• From query to write: SQL editing, editable result sets, table design, import/export, backup/restore and ER diagrams.
• Native architecture, performance first: built with Rust + GPUI, without Electron or a WebView, with low overhead and responsive interaction as design goals.
• AI with explicit boundaries: bounded reads, authorized editor edits and approval for write or high-risk tools.

Core capabilities

SQL and data

• Syntax highlighting, code folding, keyword/table/column/function completion, SQL formatting and query history.
• Paginated browsing, filters, sorting, row details and cell details.
• Editable result sets with writes back to the database and safeguards for dangerous actions.
• Import TXT, CSV, JSON and XML.
• Export SQL, TXT, CSV, JSON and XML.
• Execute existing .sql files.

Schema and ER diagrams

• Inspect columns, indexes, foreign keys, triggers and DDL.
• Create and modify tables visually, with DDL generated for the target database dialect.
• Layout ER diagrams from real foreign keys, with optional local logical relationships.
• Local logical relationships never create or modify database foreign keys.
• Export ER diagrams as JSON, DBML, Mermaid or SVG.

MongoDB workbench

• Connection-string import with SRV discovery, browsing by database / collection / view, and connection info with topology.
• The documents tab supports filter, projection, sort, collation, paging and index hint, with card, table and JSON display modes.
• The condition builder and completion cover filter, projection, sort and aggregation pipelines; aggregation, indexes and collection info each have their own tab.
• Create, edit, duplicate and delete documents; import JSON / JSONL / CSV and export JSON / JSONL / CSV.
• Collection context menu covers pin, copy name, refresh, export, rename, import and drop; databases can create collections and be dropped.

Redis workbench

• Browse the keyspace and inspect or edit common value types, TTL and key details.
• Run commands with command history in the Workbench.
• Use the embedded Redis CLI terminal.
• Observe channel messages with Pub/Sub subscriptions.
• Connect with TLS, SSH tunnels, Sentinel, Cluster, URI or Cloud configuration.

Operations, security and updates

• Connection passwords and model API keys use controlled storage and stay out of ordinary exports, logs, event snapshots and AI context.
• In-app updates verify the signed manifest and package checksum.
• Release packages include SHA-256 checksums.
• The UI supports English and Simplified Chinese with runtime switching.

Demo

Full demo · 2 min 46 sec · MP4, about 8.6 MB

https://github.com/user-attachments/assets/f6075999-a87f-4644-9d14-dbda49a7981d

Database support

Capabilities vary by database. See the user guide for details.

AI Agent

FluxDB includes an optional AI Agent for understanding schemas, writing SQL, investigating data and generating reports. It is not an unrestricted desktop assistant.

• Schema discovery: search connections, databases, schemas, tables, views and columns without reading whole tables by default.
• Bounded reads: read only the rows or content allowed by each tool contract.
• Authorized editor actions: edit only an explicitly authorized SQL range and never execute it implicitly.
• Controlled SQL and Redis execution: write or high-risk tools such as db.run and redis.execute require approval.
• Scoped context: read only authorized attachments, page references or editor snapshots.
• BI and reports: generate report drafts, charts, tables and explanations as saveable artifacts.
• MCP and plugins: configure MCP Client/Server, model bindings, plugins, tool permissions and usage diagnostics.
• User-provided models: configure your own model service and API key. Secrets use controlled storage and do not enter ordinary logs, exports or AI context.

Download and install

Open the latest release and choose the package for your platform.

• macOS: open the DMG and drag FluxDB into Applications. The app is currently ad-hoc signed and not notarized. If macOS blocks the first launch, verify the download source and allow it in System Settings → Privacy & Security.
• Windows: run the installer and optionally create a desktop shortcut.
• Linux: targets Debian-based distributions such as Ubuntu 22.04/24.04 with X11 or Wayland. Run sudo apt install ./fluxdb_<version>_amd64.deb.

Release packages include .sha256 checksums. Windows ARM64, Linux ARM64, AppImage, Flatpak and Snap are not currently available.

MySQL, MariaDB and PostgreSQL backup/restore depend on the local mysqldump/mariadb-dump, mysql/mariadb and pg_dump/psql clients. Basic queries and editing do not require them.

After installation, choose “Add connection”, select a database type, enter the connection details and test the connection. See the user guide; it currently uses Chinese screenshots and macOS shortcuts.

Run from source

git clone https://github.com/fluxdb-alt/fluxDB.git
cd fluxDB
cargo run --locked -p fluxdb-desktop

cargo build --locked --release -p fluxdb-desktop

./scripts/package-macos.sh

Contributing

Issues, feature proposals, documentation improvements and code contributions are welcome.

• Report a bug: search existing issues first, then include the app version, operating system and architecture, database type and version, reproduction steps, expected and actual results, and sanitized logs or screenshots where useful.
• Suggest a feature: describe the use case and the problem it solves. Discuss larger changes before implementation.
• Contribute code: fork the repository, create a focused branch and explain the reason and validation results in the pull request. See AGENTS.md for development conventions and the UI style guide for interface changes (both in Chinese).

Run the local quality gate before submitting:

./scripts/check.sh

Cross-platform build differences still require CI verification before release.

AI is used extensively in the development workflow. Humans own requirements, review, validation and release approval.

Roadmap

Current priorities are real-world coverage for existing databases, stability and the cross-platform release experience. Candidate directions include:

• SQL Server follow-up: Windows integrated authentication and Azure AD, contained database users, column-level permissions, user-defined server roles, named-instance auto-discovery, and database-level physical backup/restore.
• MongoDB follow-up: database-level backup/restore, performance metrics, bulk delete and bulk writes, in-table row editing, transactions and change streams.
• More databases: evaluate ClickHouse, DuckDB, Elasticsearch/OpenSearch and others based on community demand.

The roadmap does not imply current support.

Support and sponsorship

If FluxDB is useful to you, consider starring the repository, sharing it with others, or contributing through issues, documentation and code.

You can also sponsor ongoing development and AI tooling costs.

Alipay	WeChat

Acknowledgements

• Zed: GPUI ecosystem, editor and native desktop architecture.
• GPUI Component / gpui-kit: reusable UI components.
• dbx: multi-database abstraction and connection-management design reference.
• RedisInsight: Redis visualization and interaction design reference.

License

FluxDB is licensed under the GNU GPL v3.0 or later.

See NOTICE for copyright and attribution details.