# LibreDB Studio

> Self-hosted, open-source SQL IDE covering 41 database engines in one browser tab, with OIDC SSO, RBAC, query auditing and an AI query agent — MIT licensed.

- **Category:** Database Management
- **Licence:** MIT
- **Latest version:** libredb-studio-0.1.108
- **Open source:** oss
- **Languages:** Typescript
- **Platforms:** Docker, Docker Compose, Windows, Helm, Debian, NPM, Homebrew, Snap, Flatpak
- **Tags:** PostgreSQL, MongoDB, SQLite, Redis, Clickhouse, Mcp, MySQL, Sql Editor, Databases, Elasticsearch, MariaDB, Ai Assistant, Oracle, SQL, Hacktoberfest, Helm, Audit log, Self Hosted, Mssql, Trino, Oidc sso, Multi engine, Cassandra, Good First Issue, Couchbase, Druid, Database Administration, Database Client, Monaco

## Description

LibreDB Studio is an in-browser database IDE you containerize inside your own private network, so a team opens one internal URL rather than installing a desktop client machine by machine. Forty-one engine families sit behind a single object tree, query editor and results grid: Postgres, MySQL, Oracle, Db2, SQL Server, SQLite and its server variants, DuckDB, plus document, key-value, graph, vector, time-series, search, warehouse and streaming systems (MongoDB, Redis, Couchbase, Elasticsearch, ClickHouse, Trino, Kafka, Neo4j, S3 buckets, Kubernetes), and industrial protocols such as OPC UA, Modbus and SNMP. On engines whose own dialect cannot express writes — Druid and the two search backends — those controls are reported as unavailable instead of failing. Day-to-day tooling includes the Monaco editor core with schema-aware suggestions, execution-plan diagrams, foreign-key schema diagrams, snapshot comparison that emits migration DDL, a virtualized grid with in-cell editing, pivots, charts, file and clipboard export, column profiling, ORM codegen, fake-data generation and a generated data dictionary. Access control arrives with the MIT build: identity-provider sign-in with claim-to-role mapping, admin/user roles, retained query history, TOTP and passkeys. An AI assistant answers questions, tunes statements and profiles tables through Gemini, OpenAI, Ollama or an OpenAI-compatible endpoint; with no model configured nothing leaves the network, and an optional MCP endpoint lets outside AI clients query it with minted tokens. Deployment runs from a single docker run through compose, a Helm chart, npx, Homebrew, deb/rpm, Snap, winget/Chocolatey/Scoop, Linux AppImage or Flatpak desktop builds, and one-click templates on many PaaS platforms, with credentials generated on first boot. Watch for limits: agent statement execution exists only on a handful of engines, value masking happens in the browser rather than on the server, and non-container channels need Node.js 24 or Bun.

## Links

- [Homepage](https://github.com/libredb/libredb-studio)
- [Website](https://libredb.org)
- [Docker](ghcr.io/libredb/libredb-studio)
- [GitHub repository](https://github.com/libredb/libredb-studio)
- [X (Twitter)](https://x.com/libredb)
- [LinkedIn](https://www.linkedin.com/company/libredb)
- [YouTube](https://www.youtube.com/@libredb)
- [Blog](https://libredb.org/blog/)
- [Documentation (repository docs)](https://github.com/libredb/libredb-studio/tree/main/docs)
- [DeepWiki project docs](https://deepwiki.com/libredb/libredb-studio)

---
Source: https://selfhost.directory/project/libredb-studio