sqrl

MCP数据服务

提供方 DataSQRL · 包名 sqrl

数据服务 stdio · 本地进程 未标注 社区 / 第三方 Claude DesktopCursorClineClaude Code
查看接入方式

Data Pipeline Automation Framework to build MCP servers, data APIs, and data lakes with SQL.

一句话选型结论

按资料字段整理

sqrl 属于数据服务类 MCP 服务器,stdio · 本地进程,鉴权方式未标注,可直接接入 Claude Desktop、Cursor、Cline 等,有可直接执行的安装命令。接入前先确认权限范围;涉及写入、删除或付款的操作,建议先在测试环境验证。

分类数据服务 传输方式stdio · 本地进程 鉴权未标注 部署方式本地部署
GitHub 星标 210 仓库热度,无公开仓库时显示未标注
工具数 未标注 暴露给模型的工具数量
资源 / 提示词 0 / 0 资源与提示词模板数量
维护方 社区 / 第三方 核验等级 仅资料核对

工具与能力

数据服务
数量来自来源目录的标注,具体工具清单以官方文档为准
工具数量未标注暴露给模型可调用的工具个数
资源数量0可读取的资源条目
提示词数量0内置提示词模板数量

能力标签

sqrl api data-pipeline database event-driven

支持客户端

适用场景

sqrlapidata-pipeline

安装与配置

stdio · 本地进程 未标注
客户端在本机拉起子进程,数据不出本机
git clone https://github.com/DataSQRL/sqrl && cd sqrl

这条命令只是入口,具体启动参数请按官方文档补齐,不要直接照搬到别的服务器。

  1. 1
    准备运行环境

    stdio 服务在本地起进程,需要先装好 Cross-platform,再按安装命令拉取依赖。

  2. 2
    安装服务

    复制上面的安装命令执行;命令里带占位符的位置,换成自己的密钥或路径。

  3. 3
    写入客户端配置

    在客户端的 MCP 配置里填好启动命令和参数,保存后完全退出客户端再打开,让新配置生效。

  4. 4
    验证是否可用

    连接成功后随便调一个只读工具,确认能返回结果,再接入正式流程。

典型使用场景

3 类
按分类与标签整理的常见用法,具体以官方文档为准
数据库查询

用自然语言查库、取数,不用手写 SQL。

数据同步与清洗

把多来源数据拉到一处做整理。

报表取数

按口径快速拉出统计结果。

选择建议

选之前先看这几条
按条目字段给出的判断,不替你决定用不用

值得选的地方

  • 有现成的安装命令,不用自己找入口,照着跑就能起服务。
  • 已适配 Claude Desktop、Cursor、Cline,换客户端不用重新研究配置。

先想清楚的地方

  • stdio 方式要求本地装好运行时(如 Node.js),换机器要重新装一遍。
  • 资料没标注工具数量,实际能力要装好后确认。
  • 版本更新与行为细节以官方仓库为准。

更适合

  • 需要在数据服务场景里让模型直接调用外部能力的人。
  • 已经在用 Claude Desktop、Cursor 的用户,接入路径最短。
  • 想先快速试一下、能接受在本地或自有环境跑服务的人。

不太适合

  • 完全不想在本地装环境的人。
  • 需要厂商级 SLA 和技术支持的生产环境。

同类对比

6 个同类
同分类里挑最接近的几条,字段以各自详情页为准
MCP 服务器工具数传输方式部署方式安装命令最后核验
sqrl当前 未标注 stdio · 本地进程 本地部署 有 2026-04-12
Memory 9d25f64e 9 未标注 未标注 无 2026-05-17
Getzep MCP 9 stdio · 本地进程 本地部署 有 2026-05-17
anything-llm 未标注 stdio · 本地进程 本地部署 有 2026-04-11
Prisma MCP Server 未标注 stdio · 本地进程 本地部署 有 2026-05-17
Astr Bot 未标注 stdio · 本地进程 本地部署 有 2026-05-17

服务器档案

数据服务
字段来自项目仓库与公开索引,中文说明由本站标注
提供方DataSQRL上游标注的维护方 DataSQRL
分类数据服务本站按上游分类展示
连接方式stdio · 本地进程客户端在本机拉起子进程,数据不出本机
鉴权方式未标注远程服务建议用最小权限凭据
维护方社区 / 第三方核验等级 仅资料核对
包名 / 运行时sqrl · Cross-platform本地安装时用到的包标识与运行时
部署类型本地部署上游给出的部署形态
最近核验2026-04-12上游最后核验时间

详细介绍

7.3 KB
来自项目仓库与上游目录的原文,未做删改

DataSQRL: Agentic Data Engineering Harness

DataSQRL is an open-source harness for building data engineering agents designed around human control, correctness, and safety.

It extends your coding agent of choice with a SQL compiler, validator, event-time simulator, and your own skills and policies. The result is a data engineering agent you can trust with pipelines, batch jobs, data APIs (REST, GraphQL, MCP), data products, and operational data.

Why DataSQRL

1. Human Control: SQL You Understand

Agents build faster than humans can verify. DataSQRL is designed for human control and understanding so your data pipelines and applications don't become opaque liabilities.

Everything the agent builds, from ingest to transform to store to serve, is expressed as declarative SQL scripts. It's easy to follow the logic and concise enough to review in one sitting.

  • Review meaning, not plumbing. Aggregations, units, filters, and joins sit on one screen. Reviewers can check that the pipeline does what was asked instead of skimming thousands of generated lines.
  • Run it yourself. One command runs the whole pipeline locally, API included, so you can inspect real results and experiment quickly.
  • Build abstraction layers. Extend with custom UDFs, table functions, and operators that capture your business semantics.

Agents speed up the work. You still own the logic.

2. Compiler for Correctness and Introspection

A language model is a probabilistic tool that's prone to errors and hallucinations. DataSQRL provides a compiler that spans the entire data pipeline, keeping types, names, keys, and schemas consistent across Flink, Kafka, Postgres, Iceberg, and an API layer.

  • Seams are generated. Types, schemas, connector configs, API contracts, and identifier mappings are all derived from one logical model, so the systems cannot disagree.
  • Deep relational validation. The validator walks the relational plan and catches bugs that are not visible in the query text: wrong or missing primary keys, changelogs treated as event streams, time-dependent joins, unbounded state, etc. Each error comes with a suggested fix the agent can act on.
  • Event-time replay testing. The simulator runs the real deployment artifacts and replays events at their original timestamps. Late data, out-of-order events, races between streams, idle sources, updates, and deletes become deterministic test cases instead of production incidents.

3. Safety Guardrails You Can Inspect and Extend

Before anything is deployed, the compiler analyzes the whole data flow, and it writes out readable artifacts for physical and deployment plans.

  • Deep artifacts for governance. Every compilation produces the full computation DAG with table types, inferred keys, timestamps, schemas, and engine assignments, along with every deployment asset. Use these artifacts for data lineage, impact analysis, audit, and automated policy checks.
  • Scalable and fault-tolerant by design. Deployment artifacts are compiled for robust operations with automatic index selection, partitioning, and fail-over.
  • Security by construction. API endpoints use parameterized queries, and authorization is bound to JWT claims in the SQL definition.
  • Your policies, enforced. Add custom validation rules for PII handling, naming, retention, or data residency. Every pipeline the agent builds is checked against them automatically.

Customize Your Harness

Every data organization has its own conventions, domain vocabulary, compliance requirements, and target infrastructure. DataSQRL is a toolkit you use to build an agent that honors those.

Layer What you customize
Skills How your team gathers requirements, plans, implements, tests, and deploys. Includes domain knowledge and data catalog context.
Validators & policies Custom rules for governance, security, and data quality standards.
Functions & connectors Your UDFs, source and sink connectors, and data formats.
Engines & deployment Your target infrastructure: Flink, Kafka, Postgres, Iceberg, and more, on Docker, Kubernetes, or managed cloud services. Extendable to the technologies you use.
Coding agent Your choice: Claude Code, Codex, OpenCode, Pi, etc.

Packaged into one Docker container which contains the data engineering agent that your teams, CI pipelines, and platforms can call.

What DataSQRL Can Build Autonomously

  • Streaming and batch pipelines on Flink, Kafka, and Iceberg, with CDC, temporal joins, windowed aggregations, and deduplication
  • Data APIs with GraphQL, REST, and MCP endpoints generated from SQL table functions, including authentication and authorization
  • Data products as curated, documented datasets and tables for analytics and downstream teams
  • Operational data for applications and AI agents, including vector embeddings and LLM enrichment

The compiled deployment artifacts run on proven open-source technologies that you operate on existing Kubernetes or managed cloud services.

Getting Started

The basic DataSQRL agent wraps the Pi coding agent with the DataSQRL framework and skills in one Docker image. All you need is Docker and an API key from your LLM provider. Run it in your project folder:bash docker run -e ANTHROPIC_API_KEY -it --rm -e TERM -e COLORTERM -v "$PWD":/workspace -w /workspace datasqrl/core-agent

(In PowerShell on Windows, use -v "${PWD}:/workspace". For OpenAI, Bedrock, Azure, or Vertex AI, swap in that provider's environment variables.)

Then tell the agent what you need in plain English:

> Build a pipeline that ingests our order data from Kafka in real time and serves hourly revenue per product through an API.

You get SQL scripts with tests that you can read, run, and verify.

For planning, iterative refinement, and deployment workflows, install the advanced DataSQRL agent as a plugin for Claude Code, Codex, Cursor, or GitHub Copilot. The Getting Started guide covers both agents in detail.

Next steps

  • Getting Started tutorial
  • Look at examples of what DataSQRL can build:
  • Collection of self-contained data products and APIs
  • Complex enterprise example for a fictional bank, built from a semantic data catalog
  • Learn about the architecture and motivation of the DataSQRL harness
  • Use this codebase to build your own data engineering agent

Contributing

We are building DataSQRL in the open so that every organization can run a data engineering agent that it understands, trusts, and controls.

Tell us what works and what doesn't by filing an issue or starting a discussion. Code contributions are welcome. See CONTRIBUTING.md.

常见问题

8 问
答案由本站依据条目字段整理,不引用上游原文
sqrl 怎么安装?

在客户端按下面的安装命令接入:git clone https://github.com/DataSQRL/sqrl && cd sqrl。对应的包名是 sqrl。装完在客户端配置里引用同一个包或端点。

sqrl 怎么在 Claude Desktop、Cursor、Cline 里配置?

它的传输方式是stdio · 本地进程,在客户端的 MCP 配置里填好启动命令和参数,保存后完全退出客户端再打开,配置才会生效。

sqrl 收费吗?

本站不代收费用,也没有标注具体价格。是否收费取决于服务本身和背后的接口额度,正式使用前按官方说明确认。

sqrl 需要 API Key 吗?

资料没有标注鉴权方式,是否需要密钥取决于服务本身,接入前按官方文档确认。

sqrl 支持哪些客户端?

资料标注支持 Claude Desktop、Cursor、Cline、Claude Code;其他客户端只要能配置 MCP Server,也能用同样的命令或地址接入。

sqrl 能提供哪些能力?

资料显示能力标签包括 sqrl、api、data-pipeline、database、event-driven。具体工具清单以官方文档为准。

sqrl 需要什么运行环境?

运行时为 Cross-platform;属于本地进程,需要在本机装好运行环境。

sqrl 是本地还是远程?

它是本地进程(stdio),由客户端在本机拉起。传输方式为stdio · 本地进程。

热门专题

按标签挑服务器
全部专题 →

服务器信息以作者仓库与官方文档为准;分类、传输方式与鉴权方式的中文说明为中文整理,最近核验 2026-04-12。