Augments MCP Server

Augments MCP Server

A next-generation framework documentation provider for Claude Code via Model Context Protocol (MCP). Returns types + prose + examples with context-aware formatting for any npm package — not just curated ones.

Category
访问服务器

README

Augments MCP Server

A next-generation framework documentation provider for Claude Code via Model Context Protocol (MCP). Returns types + prose + examples with context-aware formatting for any npm package — not just curated ones.

mcp-name: dev.augments/mcp

What's New in v7

Version 7.0 is the biggest upgrade yet — documentation-first search, 4 new tools, BM25 indexing, and production-grade reliability.

v6 v7
3 tools 8 tools (+ diagnostics)
Types-only search Documentation-first BM25 search
8 concept synonyms 25 concept synonym clusters
Generic version diffs Real changelog-backed breaking changes
No migration guides Cross-version migration guides
No error diagnosis Curated error patterns + GitHub Issues
No package comparison Side-by-side package comparison
No dep scanning Dependency scanner (outdated/deprecated/security)
FIFO caches LRU caches with hit/miss stats
Fixed retry delays Exponential backoff + circuit breaker

Quick Start

Claude Code

# Add the MCP server (runs locally via npx)
claude mcp add -s user augments -- npx -y @augmnt-sh/augments-mcp-server

# Verify configuration
claude mcp list

Cursor

Add to your MCP config:

{
  "mcpServers": {
    "augments": {
      "command": "npx",
      "args": ["-y", "@augmnt-sh/augments-mcp-server"]
    }
  }
}

Environment Variables

Set GITHUB_TOKEN for higher GitHub API rate limits when fetching examples and documentation:

{
  "mcpServers": {
    "augments": {
      "command": "npx",
      "args": ["-y", "@augmnt-sh/augments-mcp-server"],
      "env": {
        "GITHUB_TOKEN": "ghp_your_token_here"
      }
    }
  }
}

Usage

# Get API context with prose + examples (recommended first tool)
@augments get_api_context query="useEffect cleanup" framework="react"

# Natural language queries work great
@augments get_api_context query="how to use zustand middleware"

# Search for APIs by concept (synonym-aware)
@augments search_apis query="state management"

# Get version info with real breaking changes
@augments get_version_info framework="react" fromVersion="18" toVersion="19"

# Migration guide between versions
@augments get_migration_guide package="next" fromVersion="14" toVersion="15"

# Diagnose an error
@augments diagnose_error error="Objects are not valid as a React child" package="react"

# Compare packages side-by-side
@augments compare_packages packages=["zod", "yup", "joi"]

# Scan project dependencies
@augments scan_project_deps

Tools

Tool Description
get_api_context Primary tool. Returns API signatures, prose documentation, and code examples for any npm package. Handles natural language queries with intent detection. Now with documentation-first BM25 search.
search_apis Search for APIs across frameworks by keyword or concept. 25 synonym clusters ("state" matches useState, createStore, atom, etc).
get_version_info Get npm version info, compare versions, and detect breaking changes backed by real changelogs.
get_migration_guide New. Cross-version migration guide with breaking changes, new features, deprecations, type diffs, and official migration docs.
diagnose_error New. Diagnose errors using curated patterns, GitHub Issues search, and troubleshooting docs.
compare_packages New. Compare 2-5 npm packages: downloads, bundle size, GitHub stars, dependencies, exported APIs.
scan_project_deps New. Scan package.json for outdated, deprecated, and insecure dependencies.
diagnostics Server health: version, uptime, memory, cache stats, Node.js version.

Architecture

flowchart TD
    A["Query: 'how to use useEffect cleanup'"] --> B

    B["Intent Detection → howto<br/>Query Parser → react / useEffect"]

    B --> C["Type Fetcher<br/>CDN racing · npm metadata · @types"]
    B --> D["Doc Fetcher<br/>GitHub API · BM25 search · 25 synonym clusters"]
    B --> E["Example Extractor<br/>GitHub docs · README fallback · Auto-discovery"]

    C --> F["Type Parser<br/>Signatures · Parameters · Related types"]
    D --> G["Doc Search Engine<br/>BM25 indexing · Heading-based chunks · API boost"]

    F --> H["Intent-Driven Formatter (howto)<br/>→ Examples first, prose, brief signature<br/>→ ~500-2000 tokens, 10KB max"]
    G --> H
    E --> H

Source Structure

src/
├── cli.ts                   # stdio entry point
├── server.ts                # MCP server (8 tools + diagnostics)
├── core/                    # Core modules
│   ├── query-parser.ts      # Parse natural language → framework + concept
│   ├── type-fetcher.ts      # Fetch .d.ts + README from npm/unpkg/jsdelivr
│   ├── type-parser.ts       # Parse TypeScript, extract signatures, synonym search
│   ├── example-extractor.ts # Fetch examples from GitHub docs + auto-discovery
│   ├── version-registry.ts  # npm registry integration + changelog-backed diffs
│   ├── doc-fetcher.ts       # GitHub documentation fetcher
│   ├── doc-search.ts        # BM25 inverted index search engine
│   ├── changelog-fetcher.ts # CHANGELOG.md + GitHub Releases parser
│   ├── type-differ.ts       # .d.ts diff between versions
│   └── error-patterns.ts    # Curated error pattern database
├── tools/v4/                # MCP tools
│   ├── get-api-context.ts   # Primary tool (types + docs + examples)
│   ├── search-apis.ts       # Cross-framework API search
│   ├── get-version-info.ts  # Version comparison
│   ├── get-migration-guide.ts # Cross-version migration guides
│   ├── diagnose-error.ts    # Error diagnosis
│   ├── compare-packages.ts  # Package comparison
│   └── scan-project-deps.ts # Dependency scanner
└── utils/
    ├── logger.ts            # stderr logger
    └── lru-cache.ts         # Generic LRU cache with stats

Key Features

Documentation-First Search (New in v7)

Queries like "prisma findMany with pagination" and "zustand create store with middleware" now return real documentation content. The BM25 search engine indexes docs fetched from GitHub by heading-based chunks, with API name boosting and synonym expansion.

Concept Synonyms

25 synonym clusters cover state, form, fetch, animation, routing, auth, cache, effect, middleware, pagination, validation, testing, streaming, error handling, database, layout, modal, table, upload, realtime, deployment, i18n, SSR, component, and context patterns. Bidirectional lookup means "usestate" expands to the "state" cluster.

Intent-Aware Formatting

Intent Trigger Format
howto "how to", "example of", "guide" Examples → prose → brief signature
reference "signature", "types", "parameters" Full signature → related types → 1 example
migration "migrate", "upgrade", "breaking" Prose → signature → examples
balanced Default Signature → prose → examples

Production Reliability (New in v7)

  • Exponential backoff with jitter for npm registry retries, with 429 Retry-After support
  • CDN circuit breaker — skips repeatedly failing CDN endpoints (3 failures / 5 min)
  • GitHub rate limiting — tracks remaining quota, skips fetches when exhausted
  • LRU caches with hit/miss statistics across all cache-using modules

Coverage

Any npm Package

Every npm package is supported out of the box — no curation or configuration needed. Augments resolves documentation automatically through four layers:

  1. Documentation search — fetches real docs from GitHub repos, indexes with BM25
  2. TypeScript types — bundled ("types" in package.json) or DefinitelyTyped (@types/*)
  3. Auto-discovered docs — parses the npm repository field, finds the GitHub repo, probes docs/ directories
  4. README fallback — extracts concept-relevant code blocks and prose from README.md

This means augments works with the entire npm ecosystem (~2.5M packages), not just a curated subset.

Enhanced Results for Popular Frameworks

22 frameworks have curated doc sources for richer examples: React, Next.js, Vue, Prisma, Zod, Supabase, TanStack Query, tRPC, React Hook Form, Framer Motion, Express, Zustand, Jotai, Drizzle, SWR, Vitest, Playwright, Fastify, Hono, Solid, Svelte, Angular, Redux

Barrel Export Handling

Special sub-module resolution for: React Hook Form, TanStack Query, Zustand, Jotai, tRPC, Drizzle ORM, Next.js

Local Development

# Clone and install
git clone https://github.com/augmentscode/augments-mcp-server.git
cd augments-mcp-server
npm install

# Build with tsup
npm run build

# Run locally
npm start

# Watch mode
npm run dev

# Run tests
npm test

# Run e2e tests (real network calls)
npm run test:e2e

# Type check
npm run type-check

How Augments Compares to Context7

Aspect Context7 Augments v7
Source Parsed prose docs Types + BM25-indexed docs + README
Accuracy Docs can be wrong Types must be correct, docs supplement
Context size ~5-10KB chunks ~500-2000 tokens (intent-aware)
Coverage Manual submission Any npm package (auto-discovery)
Format One-size-fits-all Intent-aware (how-to vs reference)
Search Keyword match BM25 + 25 concept synonym clusters
Freshness Crawl schedule On-demand from npm + GitHub
Migration No Cross-version migration guides
Error help No Curated patterns + GitHub Issues
Dep scanning No Outdated/deprecated/security checks

Contributing

  1. Fork the repository
  2. Create a feature branch: git checkout -b feature/amazing-feature
  3. Make your changes
  4. Run tests: npm test
  5. Submit a pull request

License

MIT License - see LICENSE for details.

Support


Built for the Claude Code ecosystem | Version 7.0.0

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