Pare

Pare

Provides MCP servers that wrap common developer tools (git, npm, docker, etc.) returning structured JSON output, enabling AI agents to reliably interact with these tools without parsing fragile terminal text.

Category
访问服务器

README

<h1><img src="assets/logo.png" alt="" width="80" valign="middle" />  Pare</h1>

CI codecov npm Downloads TypeScript License: MIT Node.js >= 20 OpenSSF Scorecard OpenSSF Best Practices

Reliable, structured CLI output for AI agents — no more parsing fragile terminal text.

Pare provides MCP servers that wrap common developer tools (git, npm, docker, test runners, etc.) and return clean, schema-validated JSON instead of raw terminal text. Agents get typed data they can act on directly, without brittle string parsing.

The Problem

Parsing CLI output is fragile. Raw terminal text includes ANSI escape codes, decorative headers, progress bars, locale-specific formatting, and platform differences that break agent workflows in subtle ways. An agent that works fine with git status on macOS may fail on Windows because the output format changed. A test runner's summary line might shift between versions, silently breaking a regex.

Pare eliminates this entire class of errors by returning schema-validated JSON with consistent field names, regardless of platform, tool version, or locale. As a bonus, structured output is significantly smaller — agents use fewer tokens per tool call:

Tool Command Raw Tokens Pare Tokens Reduction
docker build (multi-stage, 11 steps) 373 20 95%
git log --stat (5 commits, verbose) 4,992 382 92%
npm install (487 packages, warnings) 241 41 83%
vitest run (28 tests, all pass) 196 39 80%
cargo build (2 errors, help text) 436 138 68%
pip install (9 packages, progress bars) 288 101 65%
cargo test (12 tests, 2 failures) 351 190 46%
npm audit (4 vulnerabilities) 287 185 36%

Token estimates use ~4 chars/token. The biggest savings appear on verbose commands (builds, installs, tests). For simpler tools like eslint or tsc, the main advantage is reliable structured data — agents can use typed JSON directly rather than parsing strings.

How It Works

Each Pare tool returns two outputs:

  • content — human-readable text, for MCP clients that display it
  • structuredContent — typed, schema-validated JSON, ready for agents to process

This uses MCP's structuredContent and outputSchema features to provide type-safe, validated data that agents can rely on without custom parsing.

Example: git status

Raw git output (~118 tokens):

On branch main
Your branch is ahead of 'origin/main' by 2 commits.
  (use "git push" to publish your local commits)

Changes to be committed:
  (use "git restore --staged <file>..." to unstage)
        modified:   src/index.ts
        new file:   src/utils.ts

Changes not staged for commit:
  (use "git add <file>..." to update what will be committed)
  (use "git restore <file>..." to discard changes in working directory)
        modified:   README.md

Untracked files:
  (use "git add <file>..." to include in what will be committed)
        temp.log

Pare structured output (~59 tokens):

{
  "branch": "main",
  "upstream": "origin/main",
  "ahead": 2,
  "staged": [
    { "file": "src/index.ts", "status": "modified" },
    { "file": "src/utils.ts", "status": "added" }
  ],
  "modified": ["README.md"],
  "deleted": [],
  "untracked": ["temp.log"],
  "conflicts": [],
  "clean": false
}

50% fewer tokens. Zero information lost. Fully typed. Savings scale with output verbosity — test runners and build logs see 80–92% reduction.

Available Servers (28 packages, 240 tools)

Install only the servers relevant to your stack — most projects need just 2–4. The full catalog covers a wide range of ecosystems so Pare works wherever you do.

Category Servers Tools Wraps
Version Control git, github 55 git, gh
Languages & Packages npm, python, cargo, go, deno, bun, nix, dotnet, ruby, swift, jvm 101 npm, pip, cargo, go, deno, bun, nix, dotnet, gem, swift, gradle, maven
Build, Lint & Test build, lint, test, cmake, bazel 23 tsc, esbuild, vite, webpack, eslint, prettier, biome, vitest, pytest, jest
Infrastructure docker, k8s, infra, security, remote 40 docker, kubectl, helm, terraform, ansible, trivy, ssh
Utilities search, http, make, process, db 21 ripgrep, fd, curl, make, just, psql, mysql, redis, mongosh

Tool Schemas — detailed response examples and field descriptions for every tool. See also Tool Response Examples for quick JSON samples.

Quick Setup

# 1. Configure MCP servers (non-interactive)
npx @paretools/init --client claude-code --preset web

# 2. Add agent rules to your project
#    (append to existing CLAUDE.md, or copy if new)
cat node_modules/@paretools/init/rules/CLAUDE.md >> CLAUDE.md

# 3. Restart your client session

# 4. Validate
npx @paretools/init doctor

Available presets: web, python, rust, go, jvm, dotnet, ruby, swift, mobile, devops, full

Setup Guides by Client

Claude Code Claude Desktop Cursor
VS Code / Copilot Windsurf Cline / Roo Code
OpenAI Codex Gemini CLI Zed
Continue.dev

Full Quickstart Guide — presets, ecosystem mapping, validation

Manual Configuration — config paths and formats for all clients

Agent Integration Guide — rule files, hooks, CLI-to-MCP mapping

Configuration

Tool Selection

By default, every Pare server registers all of its tools. If a server exposes tools you don't need — or you want to limit which tools are available to an agent — you can filter them with environment variables.

Per-server filter — restrict a single server's tools:

# Only register status and log in the git server
PARE_GIT_TOOLS=status,log npx @paretools/git

Universal filter — restrict tools across all servers:

# Only register these specific tools across any server
PARE_TOOLS=git:status,git:log,npm:install npx @paretools/git

Disable all tools — set the env var to an empty string:

PARE_GIT_TOOLS= npx @paretools/git   # no tools registered
Env Var Scope Format Example
PARE_TOOLS All servers server:tool,... git:status,npm:install
PARE_{SERVER}_TOOLS One server tool,... status,log,diff

Rules:

  • No env var = all tools enabled (default)
  • PARE_TOOLS (universal) takes precedence over per-server vars
  • Server names use uppercase with hyphens replaced by underscores (e.g., PARE_MY_SERVER_TOOLS)
  • Whitespace around commas is ignored

Common patterns:

# Read-only git (no push, commit, add, checkout)
PARE_GIT_TOOLS=status,log,diff,branch,show

# Minimal npm
PARE_NPM_TOOLS=install,test,run

# Only specific tools across all servers
PARE_TOOLS=git:status,git:diff,npm:install,test:run

In JSON MCP config, set via the env key:

{
  "mcpServers": {
    "pare-git": {
      "command": "npx",
      "args": ["-y", "@paretools/git"],
      "env": {
        "PARE_GIT_TOOLS": "status,log,diff,show"
      }
    }
  }
}

Troubleshooting

Issue Solution
npx not found / ENOENT on Windows Use cmd /c npx wrapper (see your client's setup guide)
Slow first start Run npx -y @paretools/git once to cache, or install globally: npm i -g @paretools/git
Node.js version error Pare requires Node.js >= 20
NVM/fnm PATH issues Use absolute path to npx: e.g., ~/.nvm/versions/node/v22/bin/npx
MCP connection timeout Set MCP_TIMEOUT=30000 for Claude Code, or increase initTimeout in client config
Too many tools filling context Use tool selection env vars to limit tools, or only install the servers you need

Contributing

Each server is a self-contained package. See CONTRIBUTING.md for the full guide.

License

MIT

推荐服务器

Baidu Map

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。

官方
精选
JavaScript
Playwright MCP Server

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。

官方
精选
TypeScript
Magic Component Platform (MCP)

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。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

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。

官方
精选
本地
TypeScript
VeyraX

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。

官方
精选
本地
graphlit-mcp-server

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。

官方
精选
TypeScript
Kagi MCP Server

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。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

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。

官方
精选