mcp-server-questdb
QuestDB MCP server connects coding agents to a running QuestDB Web Console. The agent gets tools to create notebook cells, run queries, and build charts.
README
@questdb/mcp-server-questdb
Formerly published as
@questdb/mcp-bridge. Every release from 0.3.0 on is published under both names, so existing configs keep working. On an old install, runnpx @questdb/mcp-bridge upgradeonce — it migrates your agent configs to the new name.
An MCP server that connects coding agents (Claude Code, Codex, Cursor, OpenCode …) to a running QuestDB Web Console. The agent gets tools to create notebook cells, run queries, and build charts. Every action executes in the browser against your already-established QuestDB session.
Setup
Quick setup (recommended)
The interactive wizard detects your installed coding agents and writes the bridge into each one's MCP config:
npx @questdb/mcp-server-questdb setup
It walks you through two steps:
- Pick agents: multi-select from the ones it detects (Claude Code, Codex, Cursor, OpenCode, Gemini CLI).
- Review settings: optionally override
CONSOLE_ORIGINandMCP_BRIDGE_PORT; press Enter to keep the defaults.
The wizard pins each agent's config to the bridge version that ran it. Your
QuestDB Web Console expects a specific bridge version. If you're on an older
console, run the matching version: npx @questdb/mcp-server-questdb@<version> setup. The config it writes will launch that same version. (When unsure, pair first; on a version mismatch the agent is told which version to switch to.)
Manual setup
Or add it to your MCP client's config by hand (e.g. ~/.claude/.mcp.json):
{
"mcpServers": {
"questdb": {
"command": "npx",
"args": ["-y", "@questdb/mcp-server-questdb"]
}
}
}
Environment variables
| Label | Value | Default Value | Description |
|---|---|---|---|
CONSOLE_ORIGIN |
origin URL | http://127.0.0.1:9000 |
QuestDB Web Console origin. 127.0.0.1 and localhost are interchangeable. |
MCP_BRIDGE_PORT |
1–65535 |
auto-allocated | When specified, the bridge uses a fixed port. The port is bound on the first pairing attempt, pairing fails with a bridge_bind_failed error if the port is taken. |
LOG_PATH |
file path | /tmp/questdb-mcp-bridge/<ISO-ts>-<pid>.log |
Override the log file location. |
LOG_LEVEL |
ERROR / WARN / INFO / DEBUG |
INFO |
DEBUG adds heartbeats and full tool payloads. |
Commands
Your MCP client runs the bridge for you via the config above, so you rarely invoke it by hand. When you do:
| Command | Description |
|---|---|
npx @questdb/mcp-server-questdb (no args) |
Start the bridge — same as start. |
npx @questdb/mcp-server-questdb start |
Start the bridge. |
npx @questdb/mcp-server-questdb setup |
Interactively configure the bridge for your coding agents. |
npx @questdb/mcp-server-questdb --version |
Print the version and exit. Alias: -v. |
npx @questdb/mcp-server-questdb --help |
Print this help and exit. Alias: -h. |
An unknown command exits non-zero with a short error. Pin a version with
npx @questdb/mcp-server-questdb@0.3.0 start. (Installed on your PATH, the
executable is named mcp-server-questdb.)
Pairing
Before any notebook / chart / SQL tool works, your browser has to pair with the bridge. The agent drives the flow.
When the agent needs to pair, it calls get_pairing_credentials and
shows you both:
- A one-click deep link — open it in the tab showing your Web Console.
- A WebSocket URL + token — paste into the MCP pill at the bottom of the Web Console if the deep link doesn't land in the right tab.
Either path lands you on a consent prompt. Accept it and the agent's next tool call goes through.
Each bridge run generates a fresh port and pairing token, held only in memory. On restart the old credentials stop working — the agent will surface new ones the next time it needs to pair.
Logs
The bridge writes to stderr and to a log file. Tail the newest:
tail -F "$(ls -t /tmp/questdb-mcp-bridge/*.log | head -1)"
At default INFO:
2026-05-15T12:29:27.142Z [INFO] tool_call: run_query
2026-05-15T12:29:27.318Z [INFO] tool_result: run_query ok
2026-05-15T12:29:28.011Z [ERROR] tool_result: update_cell internal_error timeout after 15000ms
At DEBUG (full payloads as continuation lines):
2026-05-15T12:29:27.142Z [INFO] tool_call: run_query
2026-05-15T12:29:27.142Z [DEBUG] args: {"query":"SELECT count() FROM trades"}
2026-05-15T12:29:27.318Z [INFO] tool_result: run_query ok
2026-05-15T12:29:27.318Z [DEBUG] content: [{"type":"text","text":"..."}]
License
Apache-2.0.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。