tsx-query
Semantic React/TSX analysis MCP server that saves 70-90% tokens by using AST to retrieve precise component usages, prop flow, and state update information for AI coding assistants.
README
tsx-query
Semantic React/TSX analysis MCP server for AI coding assistants.
Save 70-90% tokens compared to grep + file reads.
Why tsx-query?
LLMs waste tokens on grep + file reads. tsx-query uses semantic AST analysis to return precisely what you need.
| Query | grep + read | tsx-query | Savings |
|---|---|---|---|
| Find Button usages | ~15,000 tokens | ~800 tokens | 95% |
| Trace prop flow (5 levels) | ~75,000 tokens | ~3,000 tokens | 96% |
| Find all setState calls | ~30,000 tokens | ~1,500 tokens | 95% |
| Analyze re-render triggers | ~40,000 tokens | ~2,000 tokens | 95% |
Installation
npm install -g @paramhq/tsx-query
Quick Start
For Claude Code
claude mcp add tsx-query -- npx -y @paramhq/tsx-query
Or manually add to your MCP config (~/.claude/mcp.json or project .mcp.json):
{
"mcpServers": {
"tsx-query": {
"command": "npx",
"args": ["-y", "@paramhq/tsx-query"]
}
}
}
For Other MCP Clients
{
"command": "npx",
"args": ["-y", "@paramhq/tsx-query"]
}
From Source
git clone https://github.com/paramhq/tsx-query.git
cd tsx-query
npm install
npm run build
npm run mcp
High-Value Tools
1. find_component_usages
Find all JSX usages of a component with powerful filtering.
{
"projectPath": "/path/to/project",
"componentName": "Button",
"withProp": "onClick",
"hasInlineHandler": true
}
Filters:
withProp/withoutProp- Filter by prop existencewithPropValue- Filter by prop value (exact or regex)hasInlineHandler- Find inline arrow handlershasSpreadProps- Find spread props usageonlyConditional- Only conditional renders
Returns: File locations, props with values, parent component, code snippets.
2. trace_prop_flow
Trace a prop through the component hierarchy. Replaces reading 5-10 files manually.
{
"projectPath": "/path/to/project",
"propName": "userId",
"startFile": "src/components/UserDetails.tsx",
"direction": "up"
}
Returns: Complete prop path through components with transformations (renamed, destructured, spread).
3. trace_state_updates
Find all locations that update a state variable. Understand state flow without reading handlers.
{
"projectPath": "/path/to/project",
"filePath": "src/components/UserForm.tsx",
"stateName": "formData"
}
Returns: All setState calls with context (handler, useEffect, inline), trigger events, async/conditional flags.
4. find_render_triggers
Analyze what causes a component to re-render with optimization suggestions.
{
"projectPath": "/path/to/project",
"filePath": "src/components/UserProfile.tsx",
"componentName": "UserProfile"
}
Returns: Props, state, context, hooks that trigger renders + memoization status + suggestions.
5. analyze_imports
Analyze import/export relationships with cached import graph.
{
"projectPath": "/path/to/project",
"filePath": "src/components/Button/index.ts",
"direction": "importedBy"
}
Returns: What imports this file, unused exports, circular dependencies.
Additional Tools
| Tool | Purpose |
|---|---|
find_hook_usages |
Find all usages of a hook across codebase |
analyze_hook_deps |
Find missing/unnecessary hook dependencies |
analyze_file |
Analyze single file structure |
list_components |
List all JSX in a file |
clear_cache |
Clear project cache after external changes |
Roadmap
- [ ]
trace_event_handler- Trace callback chains from event to side effects - [ ]
trace_context_usage- Map Context providers to consumers
Why Use Over Grep?
| Grep Limitation | tsx-query Solution |
|---|---|
| Finds "Button" in comments/strings | Only finds JSX <Button> usages |
| Can't filter by prop values | withPropValue: { prop: "variant", value: "danger" } |
| Can't detect aliased imports | Handles import { Button as Btn } |
| Can't trace prop flow | Full cross-file prop tracing |
| Reads entire files | Returns only relevant data |
Built With
- ts-morph - TypeScript AST manipulation
- Model Context Protocol SDK - MCP server implementation
- Zod - Schema validation
Author
Sukhdev - @paramhq
Contributing
See CONTRIBUTING.md for guidelines.
License
MIT - see LICENSE
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。