# DryRun — Offline PostgreSQL MCP (No DB Creds)

> DryRun snapshots your PostgreSQL schema once and provides offline linting plus migration/query checks via CLI + MCP, without sharing prod credentials.

## Install

Merge the JSON below into your `.mcp.json`:

## Quick Use

```bash
brew install boringsql/boringsql/dryrun
dryrun init
dryrun import schema.json
dryrun lint
claude mcp add dryrun -- dryrun mcp-serve
```

## Intro

DryRun snapshots your PostgreSQL schema once and provides offline linting plus migration/query checks via CLI + MCP, without sharing prod credentials.

**Best for:** teams who want database-aware agents without sharing `DATABASE_URL`

**Works with:** PostgreSQL schemas (JSON snapshots), MCP clients over stdio/SSE, Homebrew/Rust installs

**Setup time:** 10-18 minutes

### Key facts (verified)

- GitHub: 27 stars · 2 forks · pushed 2026-05-12.
- License: BSD-2-Clause · owner avatar + repo URL verified via GitHub API.
- README-verified entrypoint: `brew install boringsql/boringsql/dryrun`.

## Main

- Adopt the “schema file” model: one person dumps/imports a snapshot, everyone else runs `dryrun lint` offline as shown in the README demo.

- Use the MCP server to give agents schema intelligence without credentials; README shows `claude mcp add dryrun -- dryrun mcp-serve`.

- Treat lint output as actionable checklists: README includes example warnings/errors and suggested fixes (PKs, timestamptz, FK indexes).

### Source-backed notes

- README states `dryrun` is both a CLI tool and an MCP server, intentionally separated.
- README explains the MCP server reads snapshots and exposes tools over stdio or SSE without a live DB connection.
- README includes offline demo output with violations and fixes, and documents `dryrun mcp-serve` usage.

### FAQ

- **Does it need prod DB access for the MCP server?**: No — README emphasizes snapshots so credentials stay with DBAs.
- **How do I start the MCP server?**: Use `dryrun mcp-serve` and add it via your client config (README shows Claude).
- **Can it lint offline?**: Yes — README demonstrates `dryrun lint` working entirely from a schema file.

## Source & Thanks

> Source: https://github.com/boringSQL/dryrun
> License: BSD-2-Clause
> GitHub stars: 27 · forks: 2

---

<!-- ZH -->

## Quick Use

```bash
brew install boringsql/boringsql/dryrun
dryrun init
dryrun import schema.json
dryrun lint
claude mcp add dryrun -- dryrun mcp-serve
```

## Intro

DryRun 是 PostgreSQL 的 CLI + MCP server：先快照一次 schema，之后离线提供 lint/查询校验/迁移安全检查，并可通过 stdio/SSE 暴露给 MCP 客户端，让助手无需接触生产库凭据。

**Best for:** 想让 agent “懂库”，但不想分发 `DATABASE_URL` 的团队

**Works with:** PostgreSQL schema 快照（JSON）；stdio/SSE MCP 客户端；Homebrew/Rust 安装

**Setup time:** 10-18 minutes

### Key facts (verified)

- GitHub：27 stars · 2 forks；最近更新 2026-05-12。
- 许可证：BSD-2-Clause；作者头像与仓库链接均已通过 GitHub API 复核。
- README 中核对过的入口命令：`brew install boringsql/boringsql/dryrun`。

## Main

- 按 README 的“schema 文件”思路落地：有人导出/导入快照，其他人直接离线 `dryrun lint`。

- 用 MCP server 给 agent 供“schema 大脑”，但不暴露凭据；README 给出 `claude mcp add dryrun -- dryrun mcp-serve`。

- 把 lint 输出当可执行清单：README 示例包含主键、timestamptz、FK 覆盖索引等问题与修复建议。

### Source-backed notes

- README 明确说明 `dryrun` 同时是 CLI 与 MCP server，并刻意分离。
- README 说明 MCP server 基于快照工作，可用 stdio/SSE 暴露工具且无需实时 DB 连接。
- README 提供离线 demo 的违规输出与修复建议，并写明 `dryrun mcp-serve` 用法。

### FAQ

- **MCP server 需要生产库权限吗？**：不需要。README 强调用快照避免分发凭据。
- **怎么启动 MCP server？**：执行 `dryrun mcp-serve` 并按客户端配置添加（README 以 Claude 为例）。
- **离线能 lint 吗？**：能。README 展示了仅靠 schema 文件就能 `dryrun lint`。

## Source & Thanks

> Source: https://github.com/boringSQL/dryrun
> License: BSD-2-Clause
> GitHub stars: 27 · forks: 2


---
Source: https://tokrepo.com/en/workflows/dryrun-offline-postgresql-mcp-no-db-creds
Author: MCP Hub