Token Compressor MCP
A remote MCP server that compresses long text contexts, stores compact reusable summaries, and retrieves relevant context by title/tag/query, helping reduce token usage.
README
Token Compressor MCP
Remote MCP server for Claude / Claude Cowork that compresses long context, stores compact reusable summaries, retrieves relevant context by title/tag/query, estimates token savings, and deletes stored records.
Tools
• compress_context — compress pasted text without saving it
• save_context — save raw text and a compressed summary
• retrieve_context — retrieve saved compressed context by query, tag, title, or ID
• list_contexts — list saved records
• context_stats — show estimated savings
• delete_context — delete one record
• clear_contexts — delete all records
Local setup
npm install
npm run dev
Local endpoints after start:
• MCP endpoint: http://localhost:3000/mcp
• Inspector: http://localhost:3000/inspector
Manufact deployment
Manufact / mcp-use projects support npm run dev, npm run build, npm run start, and npm run deploy. The mcp-use docs say local servers expose /mcp for MCP clients and /inspector for testing.
Recommended cloud settings:
Node version: 20+
Build command: npm install && npm run build
Start command: npm run start
MCP endpoint: https://<your-manufact-url>/mcp
Inspector: https://<your-manufact-url>/inspector
Claude Cowork usage
After deployment, add the remote MCP URL as a Claude custom connector.
Test prompts:
Use the token compressor MCP to compress this context without saving it:
[paste non-sensitive text]
Use the token compressor MCP to save this as "Test Context" with tags test, sample:
[paste non-sensitive text]
Retrieve the compressed context for Test Context.
Privacy note
This starter uses in-memory storage. Saved records disappear when the process restarts. That is intentional for a safer first version. Do not store confidential company data in a personal/free cloud server unless approved.
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。