Skip to main content
Glama
Erenshkigal111

ccg-mcp

by Erenshkigal111
OverviewSchemaSourceRelated ServersDiscussionsAdmin
Python
Remote

CCG-MCP v0.2.0

三个 stdio MCP server，让 Claude Code Desktop 通过 API 直接调用 OpenAI Codex/GPT-5、Gemini、Claude，不依赖本地 CLI，支持第三方 API 代理网关。

Server

入口

默认模型

codex-api

codex-api-mcp

gpt-5-Users\useredb6858d\projectsx

gemini-api

gemini-api-mcp

gemini-3.1-pro-preview

claude-api

claude-api-mcp

claude-sonnet-4-6

快速启动（uvx）

# 从 GitHub 直接拉起（无需 clone）
uvx --from git+https://github.com/Erenshkigal111/ccg-mcp codex-api-mcp
uvx --from git+https://github.com/Erenshkigal111/ccg-mcp gemini-api-mcp
uvx --from git+https://github.com/Erenshkigal111/ccg-mcp claude-api-mcp

Related MCP server: Collaborative MCP Proxy Server

settings.json — 直连官方 API

Windows

{
  "mcpServers": {
    "codex-api": {
      "command": "uvx",
      "args": ["--from", "git+https://github.com/Erenshkigal111/ccg-mcp", "codex-api-mcp"],
      "env": {
        "OPENAI_API_KEY": "${env:OPENAI_API_KEY}"
      }
    },
    "gemini-api": {
      "command": "uvx",
      "args": ["--from", "git+https://github.com/Erenshkigal111/ccg-mcp", "gemini-api-mcp"],
      "env": {
        "GOOGLE_API_KEY": "${env:GOOGLE_API_KEY}"
      }
    },
    "claude-api": {
      "command": "uvx",
      "args": ["--from", "git+https://github.com/Erenshkigal111/ccg-mcp", "claude-api-mcp"],
      "env": {
        "ANTHROPIC_API_KEY": "${env:ANTHROPIC_API_KEY}"
      }
    }
  }
}

macOS / Linux

{
  "mcpServers": {
    "codex-api": {
      "command": "uvx",
      "args": ["--from", "git+https://github.com/Erenshkigal111/ccg-mcp", "codex-api-mcp"],
      "env": {
        "OPENAI_API_KEY": "$OPENAI_API_KEY"
      }
    },
    "gemini-api": {
      "command": "uvx",
      "args": ["--from", "git+https://github.com/Erenshkigal111/ccg-mcp", "gemini-api-mcp"],
      "env": {
        "GOOGLE_API_KEY": "$GOOGLE_API_KEY"
      }
    },
    "claude-api": {
      "command": "uvx",
      "args": ["--from", "git+https://github.com/Erenshkigal111/ccg-mcp", "claude-api-mcp"],
      "env": {
        "ANTHROPIC_API_KEY": "$ANTHROPIC_API_KEY"
      }
    }
  }
}

settings.json — 第三方 API 代理网关

当所有 API（OpenAI、Claude、Gemini）经由同一第三方代理（如 https://hk1.augmunt.com）转发时，使用以下模板。API Key 只写系统环境变量，不硬编码进 settings.json。

Windows（第三方网关）

{
  "mcpServers": {
    "codex-api": {
      "command": "uvx",
      "args": ["--from", "git+https://github.com/Erenshkigal111/ccg-mcp", "codex-api-mcp"],
      "env": {
        "OPENAI_API_KEY": "${env:OPENAI_API_KEY}",
        "OPENAI_BASE_URL": "https://hk1.augmunt.com/v1"
      }
    },
    "gemini-api": {
      "command": "uvx",
      "args": ["--from", "git+https://github.com/Erenshkigal111/ccg-mcp", "gemini-api-mcp"],
      "env": {
        "GOOGLE_API_KEY": "${env:GOOGLE_API_KEY}",
        "GEMINI_API_ENDPOINT": "hk1.augmunt.com"
      }
    },
    "claude-api": {
      "command": "uvx",
      "args": ["--from", "git+https://github.com/Erenshkigal111/ccg-mcp", "claude-api-mcp"],
      "env": {
        "ANTHROPIC_API_KEY": "${env:ANTHROPIC_API_KEY}",
        "ANTHROPIC_BASE_URL": "https://hk1.augmunt.com"
      }
    }
  }
}

注意：OPENAI_BASE_URL 需要 /v1 后缀；GEMINI_API_ENDPOINT 只填主机名（无协议前缀）；ANTHROPIC_BASE_URL 填完整 URL（无路径后缀）。

若 Gemini 代理为 OpenAI-compatible 模式，将 gemini-api 改为使用 codex-api-mcp，并将 OPENAI_BASE_URL 指向代理，model 填 gemini-3.1-pro-preview。

代理相关环境变量一览

变量

Server

说明

OPENAI_API_KEY

Users\useredb6858d\projectsx-api

OpenAI / 代理 key

OPENAI_BASE_URL

Users\useredb6858d\projectsx-api

自定义端点，如 https://proxy.example.com/v1（默认 OpenAI 官方）

GOOGLE_API_KEY

gemini-api

Gemini key

GEMINI_API_ENDPOINT

gemini-api

自定义 API 主机名，如 hk1.augmunt.com（默认 Google 官方）

ANTHROPIC_API_KEY

claude-api

Anthropic / 代理 key

ANTHROPIC_BASE_URL

claude-api

自定义端点，如 https://proxy.example.com（默认 Anthropic 官方）

CCG_MCP_TIMEOUT

全部

超时秒数，默认 600

Tool schema

codex-api.exec

exec(
  prompt: str,
  role: str = "",
  model: str = "gpt-5-codex",
  reasoning_effort: str = "medium",
  workdir: str = ""
) -> str

gemini-api.exec

exec(
  prompt: str,
  role: str = "",
  model: str = "gemini-3.1-pro-preview",
  workdir: str = ""
) -> str

claude-api.exec

exec(
  prompt: str,
  role: str = "",
  model: str = "claude-sonnet-4-6",
  workdir: str = ""
) -> str

烟测

git clone https://github.com/Erenshkigal111/ccg-mcp
cd ccg-mcp
pip install -e .
python tests/smoke_test.py

已知限制

  • 速率限制：API 返回 429 时，server 以 JSON 字符串返回错误，调用方决定重试。

  • 模型可用性：默认模型取决于账号权限和代理供应商发布状态；入参 model 可随时覆盖。

  • Gemini 原生 SDK：GEMINI_API_ENDPOINT 只支持 google-genai 原生协议；若代理为 OpenAI-compatible，改用 codex-api-mcp + OPENAI_BASE_URL。

  • 文件访问：三个 server 均不直接读取本地工作区文件，仅调用远程 API。

This server cannot be deployed

One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

Maintenance

ActivityInactive
ResponsivenessNo issues

Resources

Related MCP Connectors

Related MCP Servers