req-mcp-server
A Bun-based MCP server providing a persistent HTTP request toolset for testing and inspecting JSON APIs. It supports session-aware operations, including cookie persistence, automatic bearer token reuse, and configurable base URLs.
README
req-mcp-server
req-mcp-server is a Bun-based MCP server that provides a persistent HTTP request toolset for testing and inspecting APIs.
It exposes a small set of MCP tools for setting a base URL, sending requests, reusing bearer auth across calls, and resetting request session state.
What it does
- serves an MCP endpoint over HTTP using Hono
- maintains per-session request state
- persists cookies within a session
- can automatically capture and reuse bearer tokens from login responses
- supports JSON request bodies and custom headers for standard API workflows
Exposed tools
The server registers four MCP tools:
new_request_session— clears cookies and stored authget_request_base— returns the current API base URLset_request_base— sets the API base URL used for later callssend_request— sends an HTTP request through the active session
send_request is the main operational tool. The other three manage session state.
Runtime behavior
- MCP transport endpoint:
/mcp - health endpoint:
/health - default port:
9797 - configurable via
MCP_PORT
Each MCP session gets its own request client. Cookies are preserved across requests in that session, and bearer auth is reused when a successful response body includes an access_token field.
Requirements
Install
bun install
Run locally
bun run src/mcp.ts
Once started, the server is available at:
http://localhost:9797/healthhttp://localhost:9797/mcp
If you want a different port:
MCP_PORT=8080 bun run src/mcp.ts
Docker
Build and run with Docker Compose:
docker compose up --build
The container exposes port 9797 and starts the same MCP server entrypoint.
Request model
send_request accepts:
endpoint— request path such as/api/auth/memethod— HTTP method such asGETorPOSTuseStoredAuth— whether to attach the stored bearer tokenheaders— optional string-to-string headersbody— optional JSON object body
Current behavior is designed for standard JSON API traffic. Multipart uploads, binary payloads, and non-object request bodies are not the primary target of this server.
Project structure
src/
mcp.ts MCP server and HTTP transport
xrest.ts session-aware request client
utils.ts small string utility helpers
skills/
request-api-tool/ companion skill guidance for using the MCP tools
Development notes
- built with Bun and TypeScript
- HTTP server implemented with Hono
- request execution uses
impit - cookie persistence uses
tough-cookie
Health check
curl http://localhost:9797/health
Expected response:
{"status":"ok"}
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。