M
MCP Directory CURATED DIRECTORY
911 MCP servers indexed
中文
← Back to directory
D

Dbx

Official
Query and manage 70+ databases from AI agents via connections already configured in DBX
How to install ↓ Compare GitHub source repository ↗
Category
Database #25 of 58
Stars
★ 28.9k Very popular
Transport
stdio (local process)
Runtime
Node.js · Prebuilt binary
Credentials
No credential needed
License
Apache-2.0
Last commit
57FMRS · C

DBX is an actively maintained, Apache-2.0 licensed open-source project recognized on Trendshift, HelloGitHub, and Product Hunt. Its MCP server ships as a separate Rust-backed npm package with configurable read-only/read-write/full-access modes, but it depends on database connections already configured in DBX, and enabling write access requires the operator to independently assess the data risk.

Strongest · Maintenance 14/20 Weakest · Reliability 8/20

Reliability
8/20
Security and permissions
10/20
Maintenance
14/20
Documentation
13/20
Setup experience
12/20
Why each score
Reliability 8/20
The manifest declares a stdio transport and the npm package @dbx-app/mcp-server, and the README provides a runnable npx launch command plus a working .mcp.json example, with npm install said to wrap the same Rust binary via a small launcher — plausible on its face. However, none of the provided material includes CI workflows or test files that would verify the init/tool-list handshake actually completes reproducibly, and no concrete tool names/parameter schemas are enumerated (only a general description of 'listing connections, browsing tables, executing SQL'). The MCP server also depends on connections already configured in the separate DBX desktop/web app, adding an external dependency and fragility. The described compatibility shim around the legacy DBX_MCP_ALLOW_WRITES variable versus a newer centralized policy suggests a nontrivial permission state machine that could misbehave across upgrades. Per the static calibration rule, without verifiable CI+test evidence reliability is capped at 12, and it is marked down further given the thin evidence.
Security and permissions 10/20
The server offers three permission tiers (read_only / safe_write / high_risk_write) and a connection allowlist, showing intent toward least privilege — a positive. However, the material does not describe any protocol-level confirmation step specifically for high_risk_write operations invoked by an external AI agent over MCP (the 'confirmation dialogs for destructive operations' mentioned in the Features section appears to target the desktop UI, not the MCP call path). In Web/Docker deployments, DBX_WEB_PASSWORD is placed in plaintext inside the MCP client config's env block, a weak credential-handling pattern. The 'no telemetry' claim is self-reported and unverifiable from the supplied material. These factors combine for a below-midpoint score, though nothing here rises to a red-line violation (no unmitigated destructive default without any confirmation, no malware, no real secret embedded in examples).
Maintenance 14/20
The repository is Apache-2.0 licensed, not archived, and the 0.4.55 version number suggests a fairly active release cadence; it also maintains multiple active community channels (Discord, QQ, WeChat) and references a CONTRIBUTING.md, indicating reasonable governance. Star count is high (13,304) but open issues are also substantial (1,149), and no verifiable evidence of issue response time or recent commit activity is present in this material; no SECURITY.md or explicit security-response channel is mentioned either. This nets out to an active-but-partially-undocumented-governance mid-high score.
Documentation 13/20
The README's MCP section is fairly thorough, covering install commands, an mcp.json example, environment variables for multiple deployment scenarios (DBX_DATA_DIR, DBX_WEB_URL, DBX_WEB_PASSWORD, the legacy DBX_MCP_ALLOW_WRITES), and links to a dedicated MCP server README and an examples directory — the shape of layered documentation. However, the actual content of packages/mcp-server/README.md was not included in this review's material, so tool parameters, limits, cost, or troubleshooting guidance cannot be verified. The permission-mode explanation also stays conceptual without worked operational examples. This nets a mid-level score.
Setup experience 12/20
The stated install path is simple — one npx command plus a standard .mcp.json snippet — and claims compatibility with Claude Code, Cursor, and Windsurf, with precompiled native binaries offered as a Node-free alternative, which is good design for ease of use. However, real usage requires the user to already have DBX desktop or web installed and its database connections configured, an external prerequisite not elaborated within the MCP section itself; Windows portable builds need extra DBX_DATA_DIR configuration, and Web/Docker deployments need an additional password environment variable, adding cross-scenario complexity. With no verifiable automated install test or CI evidence, the static calibration rule caps setup at 15, and it is marked down further for the additional configuration burden described above.

Static review · not runListed 2026-08-06

Read the FMRS scoring method →

Fit and risk

What it can accessConnects to a database

Best for

  • Developers already using DBX desktop, web, or Docker with connections configured
  • Teams working across many different database engines who want one consistent MCP access point
  • Users who want a lightweight (20 MB) client without Java or Python runtime dependencies

Not for

  • Users without an existing DBX installation or configured connections, since the MCP server itself does not manage new database credentials — it relies on connections already set up in DBX
  • Users who specifically need an MCP server maintained by a database vendor (e.g. the MySQL, PostgreSQL, or MongoDB projects themselves) rather than DBX's own client wrapper
  • Environments with strict no-write automation requirements where operators cannot verify the current DBX centralized MCP access-mode policy

Required permissions

  • Access to database connection credentials already configured in DBX
  • Ability to execute SQL according to the mode selected in DBX Settings → MCP (read_only / safe_write / high_risk_write)
  • DBX_WEB_URL (and DBX_WEB_PASSWORD if password login is enabled) for Web/Docker deployments
  • DBX_DATA_DIR pointing to the data directory containing dbx.db for Windows portable builds
  • Node.js >= 18 when run via npx (or use the Node-free precompiled native binary instead)

Risks and side effects

  • If the access mode is set to Data read/write or Full access, the AI agent can modify or delete data in connected databases
  • Insufficiently reviewed AI-generated SQL could execute unintended write operations against production databases
  • The legacy DBX_MCP_ALLOW_WRITES=0 environment variable only enforces read-only behavior until a central MCP policy has been saved for the first time — after that it no longer applies, so operators must track policy state
  • DBX_WEB_PASSWORD is stored in plaintext in client configuration, so config files should be protected

Setup

Before you start

Runtime：Node.js · Prebuilt binary

DBX_WEB_PASSWORD optionalsecret The Web login password; set to the same value if the Web login page requires one.
Other optional settings (3)
DBX_WEB_URL optional URL of the DBX Web/Docker backend API, e.g. http://localhost:4224; only needed for Web/Docker deployments.
DBX_DATA_DIR optional Windows portable builds must point to the data directory next to DBX.exe (the folder containing dbx.db).
DBX_MCP_ALLOW_WRITES optional Legacy compatibility variable; 0 or false keeps read-only restriction and can never enable writes.
  1. Make sure DBX (desktop, web, or Docker) is installed and has at least one database connection already configured.
  2. Add the server to your .mcp.json: run via npx -y @dbx-app/mcp-server (stdio transport), or download a precompiled native binary for your platform from GitHub Releases (no Node.js required).
  3. In DBX Settings → MCP, manage the connection allowlist and choose an access mode: Read only (read_only), Data read/write (safe_write), or Full access (high_risk_write).
  4. For DBX Web or Docker deployments, set DBX_WEB_URL to point at the web backend API; if the web login requires a password, also set DBX_WEB_PASSWORD.
  5. On Windows portable builds, set DBX_DATA_DIR in the MCP config to the data directory next to DBX.exe (the folder containing dbx.db).
.mcp.json 
{"mcpServers": {"dbx": {"command": "npx", "args": ["-y", "@dbx-app/mcp-server"]}}}

Shown for Claude Code. Other clients may use a different file or key (VS Code uses "servers") — the configurator below converts it.

.vscode/mcp.json
{
  "servers": {
    "dbx": {
      "command": "npx",
      "args": [
        "-y",
        "@dbx-app/mcp-server"
      ]
    }
  }
}

Goes in your project's .vscode/mcp.json (VS Code uses a "servers" key).

Terminal
claude mcp add dbx -- npx -y @dbx-app/mcp-server

Run it in a terminal; replace any <…> placeholders with your own values first.

Generate and validate client config

Check that it works

Confirm the DBX tools (list connections, browse tables, execute SQL) appear in your client's tool list, or ask it to run "select 1" to prove the database connection works.

Troubleshooting

  1. Confirm DBX itself is installed and has at least one configured database connection — the MCP server is distributed separately and is not auto-installed with the DBX app
  2. If launching via npx fails, verify Node.js >= 18 is installed, or switch to a precompiled native binary from GitHub Releases
  3. On Windows portable builds, if data isn't found, check that DBX_DATA_DIR is set and points to the data directory next to DBX.exe containing dbx.db
  4. If connecting to a Web/Docker DBX backend fails, verify DBX_WEB_URL is correct and that DBX_WEB_PASSWORD matches the web login password when one is required
  5. If write operations are unexpectedly allowed or blocked, check the access mode (read_only / safe_write / high_risk_write) and connection allowlist in DBX Settings → MCP

Things to try

Once connected, you can ask your AI assistant things like:

  • List all database connections configured in DBX
  • Browse the tables in my mydb connection
  • Run select 1 on the local connection
  • Open the users table in DBX's UI

Use cases

Query and browse databases already configured in DBX directly from Claude Code, Cursor, Windsurf, or other MCP-compatible editors without switching tools
Have an AI agent list available connections, browse table schemas, run SQL, and jump straight to the matching table in DBX's UI
Perform exploratory queries and lightweight data operations across a multi-database environment covering 70+ engines through one consistent MCP entry point

Supported clients

Claude Code
Cursor
Windsurf

Listed from the project's documentation, not tested by this site.

Overview

DBX is a 20 MB cross-platform database client supporting MySQL, PostgreSQL, SQLite, Redis, MongoDB, DuckDB, SQL Server, Dameng, and 70+ other databases. Its MCP Server is a separate Rust-powered npm package (@dbx-app/mcp-server) that runs over stdio, letting AI coding agents such as Claude Code, Cursor, and Windsurf reuse database connections already set up in DBX's desktop, web, or Docker deployments to list connections, browse tables, execute SQL, and open tables directly in DBX's UI. The MCP server is distributed independently from the desktop application, so installing DBX does not automatically install the MCP executable.

Similar servers

Compare with other MCP servers

Source revision d4951da826f9 Data synced 2026-11-04 Read the FMRS scoring method