Token导航 LogoToken导航

Agent Skill · 研究检索

mcp-openapi-proxy(MCP OpenAPI proxy 搜索)

用于辅助 API 设计、接口文档、请求响应结构和服务集成说明。它适合让 Agent 梳理 endpoint、生成 OpenAPI 草稿、检查字段命名、整理错误码或辅助前后端联调。使用时需要确认真实业务语义、鉴权方式、分页和错误处理规则;涉及生成接口文档时,应避免凭空补字段,最好从现有代码、schema 或接口样例中提取事实。

mcp-openapi-proxy 是研究检索类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为接触敏感数据;许可证信息还不明确,第三方扫描有提醒项,按安装量在研究检索类的 73,675 个 Skill 里排第 73,543。

总安装 1周安装 12GitHub Stars 1下载 97核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
unknown

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:mcp-openapi-proxy(MCP OpenAPI proxy 搜索)
来源仓库:https://github.com/rendis/mcp-openapi-proxy
仓库路径:skills/mcp-openapi-proxy
安装命令:
npx skills add https://github.com/rendis/mcp-openapi-proxy --skill mcp-openapi-proxy
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/rendis/mcp-openapi-proxy --skill mcp-openapi-proxy

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

用于辅助 API 设计、接口文档和 OpenAPI 草稿生成。它能梳理 endpoint、检查字段命名或整理错误码。使用时需确认真实业务语义、鉴权方式和分页规则。mcp-openapi-proxy 属于研究检索类 Skill,可作为该场景下的辅助能力补充。涉及生成接口文档时应避免凭空补字段,优先提取现有事实。

建议结合代码、schema 或接口样例进行校验。

来源仓库内路径:skills/mcp-openapi-proxy

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/rendis/mcp-openapi-proxy --skill mcp-openapi-p…;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要接触敏感数据,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

mcp-openapi-proxy 怎么安装?

在终端执行 npx skills add https://github.com/rendis/mcp-openapi-proxy --skill mcp-openapi-proxy。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

mcp-openapi-proxy 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

mcp-openapi-proxy 需要什么权限?

权限级别是接触敏感数据。该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

mcp-openapi-proxy 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

mcp-openapi-proxy 用的人多吗?

公开数据是累计安装 1 次,近一周安装 12 次,下载 97 次,GitHub 1 星。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

mcp-openapi-proxy 安全吗?

第三方扫描有提醒项,说明存在需要人工确认的行为。安装前请打开源码看清它读了什么、联网到哪里。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。

先想清楚的地方

  • 需要接触敏感数据,装到常用环境前要先评估影响面。
  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 第三方扫描存在提醒项或未通过项,需要人工逐条确认。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。

不太适合

  • 环境里有大量密钥、凭证,不方便给第三方 Skill 读取的人。
  • 需要明确授权才能用到公司项目里的场景。
  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
mcp-openapi-proxy

MCP OpenAPI proxy 搜索 · 本页

11接触敏感数据未明确有提醒

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

mcp-openapi-proxy

Turns any OpenAPI 3.x spec into a lightweight MCP stdio navigator/executor. The server always exposes 3 MCP tools and uses toolName identifiers to refer to individual endpoints.

For exhaustive reference, see the repository README.md. This skill is the short operational guide for installing the binary, wiring it into an MCP client, and using the current tool contract correctly.

Always discover the actual registered tools first; never assume semantic tool names.

Install

go install github.com/rendis/mcp-openapi-proxy/cmd/mcp-openapi-proxy@latest

Quick Start

MCP_SPEC=./openapi.yaml \
MCP_BASE_URL=https://api.example.com \
MCP_AUTH_TOKEN=your-token \
mcp-openapi-proxy

MCP_BASE_URL is optional when the OpenAPI spec resolves to a single absolute server.

Recommended setup flow:

  1. Install mcp-openapi-proxy
  2. Configure your MCP client:

- Claude Code: create .mcp.json from .mcp.json.example - Codex global: run codex mcp add... - Codex project-local: create ./.codex/config.toml manually

  1. Choose auth:

- Static token: add MCP_AUTH_TOKEN - OIDC: add MCP_OIDC_ISSUER and MCP_OIDC_CLIENT_ID, then run one of: - mcp-openapi-proxy login - mcp-openapi-proxy login <mcp_name> - mcp-openapi-proxy login --mcp-config./path/to/.mcp.json - mcp-openapi-proxy login --mcp-config./path/to/.mcp.json --server <mcp_name> - mcp-openapi-proxy login --codex-server <name> - mcp-openapi-proxy login --codex-config ~/.codex/config.toml - mcp-openapi-proxy login --codex-config ~/.codex/config.toml --server <name>

  1. Start the MCP client
  2. Use list_endpoints, describe_endpoint, and call_endpoint

When login is invoked with .mcp.json or Codex config support, it reads the selected server’s env from that config entry. Shell env still overrides the config file when both are present.

Plain mcp-openapi-proxy login stays env-first. When shell env is not enough, it falls back to:

  1. ./.mcp.json
  2. ~/.codex/config.toml (or $CODEX_HOME/config.toml)

If the server name is omitted, login discovers eligible servers from the selected config file. It only considers direct command values that point to the mcp-openapi-proxy binary (plain name, full path, or .exe path). Wrapper commands such as go, env, docker, or shell scripts are ignored. If more than one eligible server exists, login prints the options and prompts you to choose one.

If the source spec comes from swag init, convert it before using the proxy. swag init emits Swagger 2.0, while mcp-openapi-proxy expects OpenAPI 3.x:

swag init -g cmd/api/main.go
swagger2openapi ./docs/swagger.json -o ./docs/openapi.json
MCP_SPEC=./docs/openapi.json mcp-openapi-proxy

Configuration

VariableRequiredDefaultDescription
MCP_SPECYes—Path or URL to OpenAPI 3.x spec (YAML/JSON)
MCP_BASE_URLNo—Explicit API base URL. If omitted, the proxy uses a single absolute OpenAPI server when available
MCP_TOOL_PREFIXNoapiPrefix for the 3 MCP tools and endpoint toolName identifiers
MCP_AUTH_PROFILENoMCP_TOOL_PREFIX or defaultNamespace for stored OIDC tokens
MCP_AUTH_TOKENNo—Global bearer token fallback
MCP_OIDC_ISSUERNo—OIDC issuer URL
MCP_OIDC_CLIENT_IDNo—OIDC client ID
MCP_OIDC_SCOPESNoSpec scopes or openid profile email offline_accessOverride OIDC login scopes
MCP_EXTRA_HEADERSNo—Comma-separated key:value pairs for every request
MCP_MAX_BODY_BYTESNo10485760Maximum response body size to buffer and return
MCP_ALLOW_INSECURE_HTTPNo0Allow sending credentials over non-loopback http:// URLs
MCP_EXCLUDE_DEPRECATEDNo0Skip deprecated endpoints when generating tools

MCP Client Setup

Claude Code (.mcp.json)

Generic example:

{
  "mcpServers": {
    "my-api": {
      "command": "mcp-openapi-proxy",
      "env": {
        "MCP_SPEC": "./openapi.yaml",
        "MCP_BASE_URL": "https://api.example.com",
        "MCP_TOOL_PREFIX": "myapi",
        "MCP_AUTH_PROFILE": "myapi"
      }
    }
  }
}

Static token variant:

{
  "mcpServers": {
    "my-api": {
      "command": "mcp-openapi-proxy",
      "env": {
        "MCP_SPEC": "./openapi.yaml",
        "MCP_BASE_URL": "https://api.example.com",
        "MCP_TOOL_PREFIX": "myapi",
        "MCP_AUTH_PROFILE": "myapi",
        "MCP_AUTH_TOKEN": "your-token"
      }
    }
  }
}

OIDC variant:

{
  "mcpServers": {
    "my-api": {
      "command": "mcp-openapi-proxy",
      "env": {
        "MCP_SPEC": "./openapi.yaml",
        "MCP_BASE_URL": "https://api.example.com",
        "MCP_TOOL_PREFIX": "myapi",
        "MCP_AUTH_PROFILE": "myapi",
        "MCP_OIDC_ISSUER": "https://auth.example.com/realms/myrealm",
        "MCP_OIDC_CLIENT_ID": "my-client"
      }
    }
  }
}

OIDC login from .mcp.json:

mcp-openapi-proxy login

Or select the server explicitly:

mcp-openapi-proxy login my-api

Or with an explicit config path:

mcp-openapi-proxy login --mcp-config ./path/to/.mcp.json

Or with an explicit config path plus explicit server:

mcp-openapi-proxy login --mcp-config ./path/to/.mcp.json --server my-api

OpenAI Codex (.codex/config.toml)

Global install via Codex CLI:

codex mcp add my-api \
  --env MCP_SPEC=./openapi.yaml \
  --env MCP_BASE_URL=https://api.example.com \
  --env MCP_TOOL_PREFIX=myapi \
  --env MCP_AUTH_TOKEN=your-token \
  -- mcp-openapi-proxy

OIDC variant via Codex CLI:

codex mcp add my-api \
  --env MCP_SPEC=./openapi.yaml \
  --env MCP_BASE_URL=https://api.example.com \
  --env MCP_TOOL_PREFIX=myapi \
  --env MCP_AUTH_PROFILE=myapi \
  --env MCP_OIDC_ISSUER=https://auth.example.com/realms/myrealm \
  --env MCP_OIDC_CLIENT_ID=my-client \
  -- mcp-openapi-proxy

Global Codex installs go into ~/.codex/config.toml.

Project-local Codex config is manual because the visible codex mcp add help does not expose a project-scoped install mode. Use ./.codex/config.toml when you want repo-local configuration:

[mcp_servers.my-api]
command = "mcp-openapi-proxy"

[mcp_servers.my-api.env]
MCP_SPEC = "./openapi.yaml"
MCP_BASE_URL = "https://api.example.com"
MCP_TOOL_PREFIX = "myapi"
MCP_AUTH_TOKEN = "your-token"

For this stdio server, use mcp-openapi-proxy login, not codex mcp login:

mcp-openapi-proxy login --codex-server my-api
mcp-openapi-proxy login --codex-config ~/.codex/config.toml
mcp-openapi-proxy login --codex-config ~/.codex/config.toml --server my-api
mcp-openapi-proxy login --codex-config ./.codex/config.toml --server my-api

codex mcp login is for OAuth that Codex manages itself. Here the proxy owns OIDC discovery, PKCE, and token storage.

If local ./.codex/config.toml is ignored, switch to the global ~/.codex/config.toml path or use a trusted project in Codex.

Gemini CLI (~/.gemini/settings.json)

{
  "mcpServers": {
    "my-api": {
      "command": "mcp-openapi-proxy",
      "env": {
        "MCP_SPEC": "./openapi.yaml",
        "MCP_BASE_URL": "https://api.example.com",
        "MCP_TOOL_PREFIX": "myapi",
        "MCP_AUTH_TOKEN": "your-token"
      }
    }
  }
}

Authentication

Credential resolution order is:

  1. MCP_AUTH_<SCHEME>_*
  2. global MCP_AUTH_TOKEN
  3. OIDC token cache for MCP_AUTH_PROFILE

Static Token (dev/CI)

Set MCP_AUTH_TOKEN to provide a global bearer fallback for bearer-compatible schemes.

For Claude Code, the usual place is .mcp.json -> mcpServers.<name>.env.MCP_AUTH_TOKEN.

Per-Scheme Credentials

Use the OpenAPI security scheme name, uppercased and sanitized by replacing ., -, /, and spaces with underscores.

  • http bearer, oauth2, openIdConnect → MCP_AUTH_<SCHEME>_TOKEN
  • http basic → MCP_AUTH_<SCHEME>_USERNAME and MCP_AUTH_<SCHEME>_PASSWORD
  • apiKey → MCP_AUTH_<SCHEME>_KEY

Example:

MCP_AUTH_PARTNER_AUTH_V2_TOKEN=secret-token
MCP_AUTH_ADMIN_BASIC_USERNAME=alice
MCP_AUTH_ADMIN_BASIC_PASSWORD=s3cr3t
MCP_AUTH_X_API_KEY_KEY=dev-key

OIDC PKCE (production)

Standard OIDC discovery (recommended):

MCP_OIDC_ISSUER=https://auth.example.com/realms/myrealm \
MCP_OIDC_CLIENT_ID=my-client \
mcp-openapi-proxy login

If your MCP client uses .mcp.json or Codex TOML, these forms now read the selected server’s env automatically:

  • mcp-openapi-proxy login
  • mcp-openapi-proxy login my-api
  • mcp-openapi-proxy login --mcp-config./path/to/.mcp.json
  • mcp-openapi-proxy login --mcp-config./path/to/.mcp.json --server my-api
  • mcp-openapi-proxy login --codex-server my-api
  • mcp-openapi-proxy login --codex-config ~/.codex/config.toml
  • mcp-openapi-proxy login --codex-config ~/.codex/config.toml --server my-api

Keep the same MCP_AUTH_PROFILE, MCP_OIDC_ISSUER, and MCP_OIDC_CLIENT_ID values there so the running server uses the token cache created by login.

If the server name is omitted, login only considers direct mcp-openapi-proxy commands from the selected config file. Wrapper commands such as go, env, docker, or shell scripts are not eligible for login discovery. When multiple eligible entries exist, login lists them and prompts you to choose one.

If you also set those values in the shell when invoking login, the shell values win.

Application-specific discovery:

MCP_BASE_URL=https://api.example.com mcp-openapi-proxy login

Commands:

  • mcp-openapi-proxy login
  • mcp-openapi-proxy status
  • mcp-openapi-proxy logout

Scopes come from MCP_OIDC_SCOPES when set. Otherwise, if MCP_SPEC is available, the proxy unions scopes from the OpenAPI security requirements. If neither source is available, it falls back to openid profile email offline_access.

Tokens are stored at ~/.mcp-openapi-proxy/<profile>-tokens.json, where <profile> comes from MCP_AUTH_PROFILE or falls back to the tool prefix / default.

Registered Tools and Endpoint IDs

The MCP server always exposes exactly these tools:

  • {prefix}_list_endpoints
  • {prefix}_describe_endpoint
  • {prefix}_call_endpoint

Always inspect the registered tools first. Do not assume the server exposes one MCP tool per endpoint anymore.

Each OpenAPI endpoint still has a stable toolName:

{prefix}_{method}_{sanitized_path}

That toolName is the identifier you pass into describe_endpoint and call_endpoint.

Feature Flags API example with MCP_TOOL_PREFIX=fe:

OperationEndpoint toolName
GET /admin/featuresfe_get_admin_features
POST /admin/featuresfe_post_admin_features
GET /admin/features/{key}fe_get_admin_features_key
PATCH /admin/features/{key}/togglefe_patch_admin_features_key_toggle
GET /admin/workspacesfe_get_admin_workspaces

Names like fe_list_features or fe_toggle_feature are not part of the product contract.

Input and Output Contract

Recommended agent flow:

  1. Call {prefix}_list_endpoints
  2. Pick a toolName
  3. Call {prefix}_describe_endpoint if you need the exact contract
  4. Call {prefix}_call_endpoint

list_endpoints

Input:

{
  "q": "feature",
  "path_prefix": "/admin",
  "method": "PATCH",
  "auth": "bearer",
  "limit": 25
}

Output:

{
  "items": [
    {
      "toolName": "fe_patch_admin_features_key_toggle",
      "method": "PATCH",
      "path": "/admin/features/{key}/toggle",
      "description": "Toggle a feature flag",
      "requiredAuth": "bearer",
      "tags": ["flags", "admin"],
      "deprecated": false
    }
  ],
  "nextCursor": ""
}

describe_endpoint

Input:

{
  "toolName": "fe_patch_admin_features_key_toggle"
}

Output includes:

  • toolName, method, path, summary, description, deprecated
  • requiredAuth and full securityRequirements
  • parameters.path/query/headers/cookies
  • requestBody with media types, schema, examples, and encoding
  • responses with status, description, headers, and body schemas
  • externalDocs, servers, baseURLHint

call_endpoint

Input:

{
  "toolName": "fe_patch_admin_features_key_toggle",
  "path": {
    "key": "checkout"
  },
  "headers": {
    "X-Workspace": "acme"
  },
  "body": {
    "enabled": true
  }
}

X-Workspace is sent in headers per request or configured globally with MCP_EXTRA_HEADERS. Workspace selection is header-based, not a dedicated tool.

call_endpoint returns the HTTP envelope:

{
  "status": 200,
  "content_type": "application/json",
  "headers": {
    "X-Trace": ["abc123"]
  },
  "body": {
    "id": "item-1"
  }
}
  • API 4xx/5xx responses preserve the real HTTP response and set IsError=true
  • Proxy/runtime failures return status: 0 plus proxy_error
  • Binary payloads are represented as base64 wrappers
  • Deprecated endpoints are listed by default and only excluded with MCP_EXCLUDE_DEPRECATED=1

Multiple APIs

Use distinct prefixes to run side-by-side:

{
  "mcpServers": {
    "users-api": { "command": "mcp-openapi-proxy", "env": { "MCP_TOOL_PREFIX": "users", "..." : "..." } },
    "billing-api": { "command": "mcp-openapi-proxy", "env": { "MCP_TOOL_PREFIX": "billing", "..." : "..." } }
  }
}

Common Mistakes

  • Missing MCP_SPEC → startup fails immediately
  • No usable base URL → set MCP_BASE_URL explicitly or declare a single absolute OpenAPI server
  • Spec generated by swag init → detect swagger: "2.0" or a generated swagger.json, convert it with swagger2openapi, then point MCP_SPEC at the resulting OpenAPI 3.x file
  • Missing per-scheme auth → configure MCP_AUTH_<SCHEME>_*, MCP_AUTH_TOKEN, or run mcp-openapi-proxy login
  • Spec URL unreachable → fails at startup; check network/VPN
  • Authenticated calls to non-loopback http:// → blocked unless MCP_ALLOW_INSECURE_HTTP=1
  • Treating endpoints as MCP tools → endpoints are discovered as toolName values; only list_endpoints, describe_endpoint, and call_endpoint are registered
  • Calling call_endpoint without discovery → list first, then describe if needed, then call with toolName

For exhaustive reference, examples, and troubleshooting, see the repository README.md.

平台分布

Codex

37.02%
按下载量换算36

Claude

30.91%
按下载量换算30

Cursor

19.29%
按下载量换算19

Gemini CLI

8.73%
按下载量换算8

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

可疑

权限和风险

接触敏感数据

该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。来源安全扫描存在提醒或未通过项,安装前请自行复核。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills