LXAgents MCP Server
Delivers the LXAgents shared agent instruction set via MCP prompts and resources. It includes read-only tools for setting up and checking agent instructions in repositories.
README
mcp-server
The MCP server that delivers the LXAgents shared agent instruction set. A repository connects it as a connector and reads the conventions it needs — branching, commits, pull requests, task workflow, the creators, the directory architecture — instead of cloning or vendoring a copy of them.
- Server id:
lxagents-agents-base - Package:
@lxagents/agents-base - Surface: MCP prompts and resources, plus four read-only tools for clients that only enumerate tools.
Key features
agents-setupprompt — the full setup procedure that builds a repository'sAGENTS.md,.agents/tree, wiki, and memory.check-duplicate-agents-instructionprompt — finds instructions a repository duplicates from the shared set and removes them. Runs only when asked.- 25 instruction resources under
agents://, plusagents://manifest.jsonlisting every file with a content hash — one read instead of walking the set. - Four read-only tools —
agents_setup,agents_check_duplicate_instructions,agents_list_instructions,agents_read_instruction— delivering the same content as the prompts, for clients that surface tools only. - Many clients at once. Content is loaded once into a frozen registry; each request or session gets its own server instance, so nothing is shared and any process can serve any request.
Quick start
npm install
npm test
# Local, over stdio
npm start
# Remote connector surface, over streamable HTTP
npm run start:http # http://localhost:3000/mcp
With Docker:
docker compose up --build
Connect it
Add a custom connector pointing at https://<host>/mcp, named
lxagents-agents-base. Include the /mcp path — without it the handshake fails,
and clients report that as a sign-in error rather than a wrong address. For local development over stdio, see
wiki/guides/connect-a-repository.md.
Documentation
- Overview — what this serves and why it is a server.
- Architecture — registry, transports, concurrency.
- MCP surface — every prompt and resource.
- Local setup — running and testing it.
Working with agents
The instruction set this server delivers is also the instruction set this repository
follows. Start at AGENTS.md; the canonical content lives in
content/.
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。