mcp-evidence-api
An MCP server that makes backend API requests and packages request+response pairs as evidence to prove an endpoint works.
README
mcp-evidence-api
An MCP server that makes backend/API requests and packages request+response pairs as evidence — proving an endpoint works. Sibling project to mcp-evidence, which does the same thing for browser/UI flows; this one is for testing backend behavior with no browser involved (video/trace don't mean anything without a page, so this is a separate, much simpler tool rather than an extension of that one).
Generic and reusable: no assumptions about any particular API. You pass a
baseUrl per session, and evidence is written into the consuming project's
working directory at .evidence/<featureName>/<timestamp>/.
Install
Register with Claude Code
Per-user (available in every project):
claude mcp add --scope user evidence-api -- npx -y github:meovan07/mcp-evidence-api
Or per-project, add to .mcp.json:
{
"mcpServers": {
"evidence-api": {
"command": "npx",
"args": ["-y", "github:meovan07/mcp-evidence-api"]
}
}
}
This repo is public, so no GitHub credentials are needed on the machine
running npx.
Consuming projects should add .evidence/ to their own .gitignore.
Tools
| Tool | Purpose |
|---|---|
start_evidence_session({ featureName, baseUrl? }) |
Creates the evidence dir. Returns sessionId. |
request({ sessionId, name?, method, url, headers?, body? }) |
Makes an HTTP request, logs the full request/response pair. url resolves against baseUrl if relative. |
finish_evidence_session({ sessionId, summary? }) |
Writes requests.json and manifest.json, returns request/failure counts. |
A session left open for 10 minutes with no tool calls is auto-finished. The
server also flushes any open sessions on SIGINT/SIGTERM so evidence isn't
lost if the process is killed mid-run.
Sensitive request/response headers (authorization, cookie,
set-cookie, x-api-key, api-key, x-auth-token, proxy-authorization)
are redacted before being written to disk. This only covers header names —
if an endpoint echoes a secret back inside a JSON response body, that isn't
redacted, since there's no reliable way to tell a secret-looking field from a
normal one. Don't point this at endpoints that echo credentials in response
bodies without being aware evidence files will contain them in plaintext.
Example
start_evidence_session({ featureName: "users api", baseUrl: "http://localhost:4000" })
-> sessionId, evidenceDir
request({ sessionId, name: "create user", method: "POST", url: "/users", body: { name: "Ada" } })
request({ sessionId, name: "get user", method: "GET", url: "/users/1" })
finish_evidence_session({ sessionId, summary: "Users API create+fetch works" })
Resulting evidence directory:
.evidence/users-api/2026-07-08T08-41-21-754Z/
requests.json
manifest.json
requests.json is an array of full request/response records (method, url,
headers, body, status, timing). manifest.json is a summary: request count
and how many came back non-2xx.
Development
npm install
npm run build # tsc -> dist/
npm run dev # tsc --watch
npm start # node dist/index.js
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。