DataOmni icon DataOmni v1.1.0

One native client for your SQL, document, KV, graph, and search databases.

Built with Tauri 2 and Rust — no 1.5 GB JVM, no tab caps, and zero plaintext passwords on disk. Preview every grid edit as exact SQL before commit, connect to Oracle out of the box, and give Claude Code, Codex, Cursor, or any MCP client strictly read-only access while keeping PROD invisible.

MIT open source
Free for personal & commercial teams, unlimited tabs
Zero telemetry
Direct local connections, no account or cloud sync
OS keychain only
Zero plaintext passwords or API keys on disk
Oracle OCI bundled
Connect immediately without separate driver setup
DataOmni main SQL workbench showing object explorer, multi-tab SQL editor, and paginated query results
Multi-tab SQL workbench — Dialect-aware statement splitting, live schema completion, server-side cancellation, and environment badges on every tab.

17 engines verified against real database instances:

  • PostgreSQLSQL
  • MySQLSQL
  • OracleOCI
  • SQL ServerTDS
  • SQLiteFile
  • DuckDBOLAP
  • ClickHouseHTTP
  • MariaDBSQL
  • TiDBSQL
  • OceanBaseSQL
  • CockroachDBSQL
  • openGaussPG
  • MongoDBDoc
  • RedisKV
  • Neo4jGraph
  • ElasticsearchSearch
  • OpenSearchSearch

Built around how engineers actually debug, edit, and ship data

Every workflow in DataOmni answers a concrete failure mode of traditional database GUIs — heavy runtimes, silent cell overwrites, opaque errors, and unsafe credential sharing.

Native Rust speed, with Oracle and enterprise quirks solved out of the box

Stop juggling separate tools for Postgres, Oracle, MongoDB, and Redis — or spending half an afternoon configuring Instant Client paths and SSH tunnels.

  • Single native Tauri 2 binary backed by Rust drivers; streams large DuckDB Parquet files and paginated results without hogging memory.
  • Oracle Instant Client (Basic Light) is bundled right inside the installer, alongside a built-in local relay for openGauss default sha256 accounts.
  • Connection failures tell you exactly which layer broke: DNS lookup, refused TCP port, TLS handshake, SSH known_hosts fingerprint, or credentials.
Runtime & zero-ceremony stack TAURI 2 + RUST
Desktop shellWebView
Tauri 2 + Rust + system WebView — no bundled JVM or Chromium bloat
Oracle & DuckDBBundled
Oracle Instant Client (Basic Light), DuckDB 2.0, and SQLite compiled right in
Auth & networkSSH · TLS
Built-in SSH jump-host tunnel, TLS/X.509 verification, and openGauss sha256 relay
OS KeychainKeyring
macOS Keychain, Windows Credential Manager, and Linux Secret Service only
Actionable errorsDiagnostic
Separates DNS failure, port refusal, handshake drop, SSH host-key mismatch, and bad auth

Every write previews as exact SQL — zero silent blur commits

Grid editors become dangerous when clicking away silently fires an UPDATE or overwrites a row changed by another session.

  • Edits, inserts, and deletes queue locally so you can inspect the exact parameterized SQL before anything touches the server.
  • Generated UPDATE and DELETE statements include original column values in WHERE; if affected_rows disagrees, the transaction rolls back.
  • Connections carry explicit DEV, TEST, STAGING, or PROD badges, enforcing stricter confirmation before DROP, TRUNCATE, or unguarded writes.
Queued edit preview · 1 cell modified 1 TRANSACTION
-- Previewed before execution; bound via driver parameters
BEGIN;

UPDATE "public"."products"
   SET "price" = $1          -- 489
 WHERE "id" = $2             -- 1 (primary key)
   AND "price" = $3;         -- 429 (optimistic check)

-- Verify affected_rows == 1, then:
COMMIT;
If another session changed price from 429 in the meantime, affected_rows is 0 and the entire batch rolls back.

Schema diff, visual execution plans, and deterministic ORM codegen

Move from ad-hoc queries to structural engineering without relying on black-box plugins or probabilistic AI output.

  • Compare tables, columns, primary keys, indexes, and foreign keys across two connections to generate a reviewable DDL sync script.
  • Inspect expandable execution plan trees across PostgreSQL, MySQL, TiDB, OceanBase, SQL Server, Oracle, and CockroachDB with >10x skew highlights.
  • Export whole-database ER diagrams (SVG, PNG, PDF, or Markdown dictionary) and generate Prisma, TypeORM, MikroORM, SeaORM, or sqlx models deterministically.
Schema diff & deterministic Rust / TypeScript codegen DETERMINISTIC
-- 1. Schema diff sync script (staging -> dev)
ALTER TABLE "public"."orders"
  ADD COLUMN "currency" varchar(3) NOT NULL DEFAULT 'USD';
CREATE INDEX "idx_orders_customer_created"
  ON "public"."orders" ("customer_id", "created_at");

// 2. Deterministic ER -> sqlx / Prisma / SeaORM / TypeORM
#[derive(Debug, Clone, sqlx::FromRow)]
pub struct Order {
  pub id: i64,
  pub customer_id: i64,
  pub currency: String,
  pub total_cents: i64,
}
Generated directly from live catalog metadata — 100% deterministic, offline, and repeatable.

Standard MCP & CLI for every AI coding tool, with PROD physically hidden

Connect Claude Code, Codex, Cursor, Windsurf, Copilot, Gemini CLI, or Claude Desktop directly to your configured connections and SSH tunnels — without ever pasting a password into a config file.

  • Off by default per connection. Enabling read-only access binds an HMAC capability to your OS keychain secret — editing JSON on disk cannot unlock it.
  • Production (PROD) connections are structurally invisible to both dataomni cli and dataomni mcp.
  • Every query passes two independent barriers: a dialect-specific read-only lexer gate plus a database read-only transaction that always ends in ROLLBACK.
dataomni mcp & dataomni cli PROD HIDDEN
$ dataomni cli connections
[{"name":"shop_dev","db_type":"postgres","access":"read_only"}]

$ dataomni cli query shop_dev "SELECT id, total FROM orders LIMIT 2"
{"columns":["id","total"],
 "rows":[[1,1467],[2,1032.5]],"truncated":false}

$ dataomni cli query shop_dev "DELETE FROM orders"
refused: statement is not read-only (exit code 3)

# Export full CLI schema & usage for any terminal agent
$ dataomni cli guide
Standard stdio MCP — pick your AI tool and copy:
claude mcp add dataomni -- /Applications/DataOmni.app/Contents/MacOS/dataomni mcp

One client across 18 real-server-tested databases

Every supported database is verified by automated smoke suites against real server instances (src-tauri/tests/*_smoke.rs), never assumed from wire-protocol similarity.

Relational & Analytical SQL

13 tested servers

One full-featured SQL workbench across OLTP databases, embedded analytics, and distributed SQL compatible servers.

  • PostgreSQL 16
  • MySQL 8.4 & 5.7
  • Oracle 23ai
  • SQL Server 2022
  • SQLite
  • DuckDB 2.0
  • ClickHouse
  • MariaDB
  • TiDB
  • OceanBase
  • CockroachDB
  • openGauss
  • Dialect-aware statement splitting (including SSMS GO batches and stored procedures) plus live catalog completion
  • Queued grid edits & table designer with exact DDL/DML preview and optimistic concurrency checks
  • Visual execution plan trees flagging >10x row-estimation skews, bind parameters (:name), and streaming CSV/SQL/Markdown export
  • Cross-connection schema diff sync scripts, whole-DB ER diagrams, and deterministic ORM codegen

Document, KV, Graph & Search Workspaces

5 dedicated workspaces

Purpose-built native workspaces instead of forcing non-relational engines into fake SQL tables.

  • MongoDB 8.0
  • Redis 7.4 (Standalone / Sentinel / Cluster)
  • Neo4j 5.26 / 2026
  • Elasticsearch 8 / 9
  • OpenSearch 3
  • MongoDB: mongosh filter/sort/projection bar, document edits, aggregation pipeline runner, index inspector, and db.runCommand console
  • Redis: non-blocking SCAN across cluster shards, editors for all 6 types, automatic JSON & Python pickle decoding, and redis-cli console
  • Neo4j: Bolt 5 Cypher workbench, interactive node/relationship graph canvas, visual property editing, and EXPLAIN / PROFILE plans
  • Elasticsearch & OpenSearch: Kibana Dev Tools request console, ES|QL & SQL tabular views, mapping viewer, and _seq_no optimistic-lock edits
Inspect full 18-server smoke-test compatibility ledger (versions, test counts & known gaps) Counts show passed / skipped / fixture-unsupported
Server Tested Connection type Status Smoke (P/S/F) Verified behavior & known gaps
PostgreSQL 16 PostgreSQL Supported 39/0/0 Full SQL workbench, actual execution plans (EXPLAIN ANALYZE), bind parameters, schema diff, pg_dump backup.
MySQL 8.4 MySQL Supported 42/0/0 Multi-result CALL support, JSON editor, TLS verification, schema diff, mysqldump backup.
MySQL 5.7 5.7.44 MySQL Usable, with gaps 27+3/0/0* TLS must be set to Disable (no ECDHE ciphers in 5.7); column renames restate definitions via CHANGE COLUMN.
SQLite Bundled SQLite Supported 18/0/0 Built in, Windows >260-char paths supported, VACUUM INTO live backup, schema diff and ORM codegen.
MariaDB 11.4 MySQL Supported 43/0/0 Dedicated preset in connection form; JSON columns are LONGTEXT aliases and edit as text.
TiDB 8.5 MySQL Supported 40/0/2* Parses EXPLAIN FORMAT='tidb_json'; warns when DROP is non-transactional; splits combined column + table renames.
SQL Server 2022 SQL Server Supported 33/0/0 SSMS-compatible GO batching, multiple result sets, estimated & actual XML execution plans, schema diff & codegen.
Oracle 23ai Free (23.26) Oracle (bundled OCI) Supported 29/1/0 Instant Client bundled; connect by Service Name or SID; server-side cancel; use SSH tunnel when encryption is needed.
DuckDB 2.0.0 (built in) DuckDB Supported 19/0/0 Streams Parquet/JSON without loading whole files into RAM; Windows long paths supported; EXPORT DATABASE backup.
OceanBase 4.4.2 CE (MySQL) MySQL Usable, with gaps 39/0/3* Parses OceanBase CHILD_n JSON plans; inspect SQL preview before altering columns that use expression defaults.
CockroachDB 25.2 PostgreSQL Usable, with gaps 39/0/0 Reads triggers via SHOW CREATE TRIGGER; parses EXPLAIN (VERBOSE) & EXPLAIN ANALYZE; runs schema changes per statement.
openGauss 5.0.3 (PG mode) PostgreSQL Usable, with gaps 40/0/0 Local protocol relay authenticates default sha256-password users seamlessly over configured TLS.
ClickHouse 25.8 ClickHouse (HTTP) Usable, with gaps 17/0/0 HTTP/HTTPS interface; single-row MergeTree edits pin _part block + _part_offset because ClickHouse has no row locks.
MongoDB 8.0 MongoDB Usable, with gaps 34/0/0 Dedicated workspace with mongosh filter/edit syntax, aggregations, db.runCommand console, SRV (Atlas), TLS & X.509.
Redis 7.4 Redis Supported 14/0/0 Standalone, Sentinel, and Cluster; SCAN browsing, 6 data type editors, JSON & pickle decoding, redis-cli console.
Neo4j 5.26 LTS / 2026.09 Neo4j (Bolt 5) Usable, with gaps 10/0/0 Cypher editor, interactive graph view, visual node & relationship editing, EXPLAIN and PROFILE plans.
Elasticsearch 8.19 / 9.5 HTTP / API Key Usable, with gaps 7/0/0 Kibana Dev Tools console, hits/ES|QL/SQL/aggregation tables, mapping viewer, optimistic-concurrency document edits.
OpenSearch 3.8 Elasticsearch Usable, with gaps 6/1/0 Shares the Elasticsearch workspace over basic auth + TLS (no API key auth; read-only accounts cannot list tree).

How DataOmni compares

An honest look at architectural trade-offs across desktop database clients.

Dimension DataOmni JVM Suites (e.g. DBeaver) Commercial Native (e.g. TablePlus / Navicat)
Runtime & memory Java / Eclipse RCP; higher idle RAM, very broad JDBC plugin catalog Native binaries with fast startup
License & limits Apache 2.0 CE (NoSQL & enterprise features require paid PRO edition) Paid per-seat license or subscription (free tier caps open tabs/connections)
SQL + NoSQL in one app CE covers SQL; MongoDB, Redis & graph workspaces require PRO Varies by product edition or paid add-on tier
Oracle & openGauss setup Downloads Maven JDBC jars on first connection Bundled or manual local OCI path configuration
Grid write safety Manual transaction mode & script preview available Staged change preview; optimistic row checks vary by driver
Credential storage Local encrypted file by default; OS keyring via extra integration OS keychain or vendor-encrypted local profile
AI Agent CLI & MCP Requires external MCP server with separate DB credentials Requires external MCP server with separate DB credentials

Where others still win: if your workflow depends on DB2, Teradata, Snowflake, or dozens of legacy JDBC drivers, DBeaver or long-standing commercial suites offer broader long-tail coverage. If your daily stack is PostgreSQL, MySQL, Oracle, SQL Server, SQLite, DuckDB, ClickHouse, and MongoDB / Redis / Neo4j / Elasticsearch — and you want native speed, explicit SQL guardrails, MIT licensing, and safe local MCP — DataOmni is purpose-built for you.

Frequently asked questions

Straight answers on licensing, credential security, code signing, and agent safety.

Is DataOmni really free for commercial work?

Yes. DataOmni is released under the permissive MIT License. Individuals, startups, and enterprise teams can use it without seat fees, mandatory accounts, or artificial caps on open tabs and saved connections.

Does DataOmni collect telemetry or upload my queries, schemas, or passwords?

No. Connections go straight from your machine (or your configured SSH jump host) to your database. Passwords, SSH key passphrases, and AI API keys live exclusively in your OS keychain (macOS Keychain, Windows Credential Manager, or Linux Secret Service) — never in plaintext JSON. The only background network call is a once-per-day GitHub Releases version check, which you can disable in Settings → About & Diagnostics. AI features are off by default and can even be compiled out via Cargo's --no-default-features for air-gapped intranets.

Why does Oracle work out of the box without installing Instant Client manually?

Every official DataOmni installer bundles Oracle Instant Client (Basic Light) under Oracle's redistributable license (included in the package). On macOS and Windows you just enter your host and Service Name or SID; on Linux, .deb and .rpm packages declare the standard system libaio dependency.

Why do macOS Gatekeeper or Windows SmartScreen warn on the first launch?

Installers are built reproducibly on GitHub Actions from public source, and the project has not yet purchased an Apple Developer ID or Windows EV code-signing certificate. On macOS, click "Open Anyway" in System Settings → Privacy & Security (or run xattr -dr com.apple.quarantine /Applications/DataOmni.app once). On Windows, click "More info" → "Run anyway". Subsequent launches open normally.

Which AI tools work with DataOmni MCP & CLI, and how is PROD protected?

DataOmni implements standard stdio MCP (JSON-RPC 2.0) alongside structured CLI JSON output, so it works out of the box with Claude Code, Codex, Cursor, Windsurf, GitHub Copilot, Gemini CLI, Zed, and Claude Desktop. Agent access is off by default on every connection; enabling read-only access binds an HMAC capability to your keychain secret so editing connections.json cannot bypass it. Production (PROD) connections are unconditionally hidden, and every query must pass both a dialect-specific read-only lexer gate and a database read-only transaction ending in ROLLBACK.

Will upgrading (or rolling back) wipe my saved connections, open tabs, or query history?

No. DataOmni enforces an automated on-disk compatibility gate (fixtures/on-disk-0.5): unknown future fields or driver types in connections.json are preserved verbatim on save, and older builds refuse to overwrite workspace or history snapshots written by a newer schema version.

Download DataOmni v1.1.0

Oracle Instant Client (Basic Light) is bundled in every package. Because builds do not yet carry paid Apple Developer ID or Windows EV certificates, the first launch requires a one-time OS confirmation.

macOS (Apple Silicon)

macOS 12+ · arm64 (.dmg)

On first open, click "Open Anyway" in System Settings → Privacy & Security, or clear the quarantine attribute:

xattr -dr com.apple.quarantine /Applications/DataOmni.app

Windows (x64)

Windows 10/11 · per-user NSIS (.exe) or machine-wide MSI

On first launch, click "More info" → "Run anyway" in Windows SmartScreen. ARM64 Windows runs the x64 build under emulation.

Linux (x64)

glibc 2.34+ (Ubuntu 22.04+, Fedora 42+) · .deb & .rpm

Saving passwords requires an unlocked Secret Service keyring (such as gnome-keyring). Install via apt or dnf:

sudo apt install ./DataOmni_1.1.0_amd64.deb