quota-dashboard-mcp
Queries real-time AI subscription quota for Claude Code Max, Kimi, and Z.ai. Tokens are read from environment variables and never stored.
README
quota-dashboard-mcp
A local-run, privacy-first MCP server that exposes real-time AI subscription quota for Claude Code Max, Kimi, and Z.ai. It uses stdio transport, so it works with Claude Code, Cursor, VS Code, and any other MCP stdio client.
Tokens stay on your machine: they are read from environment variables at call time, never persisted to disk, and never sent anywhere except the provider's own API.
Looking for a GUI? See the sibling project
quota-dashboard.
Tools
| Tool | Description |
|---|---|
get_quota_summary |
Unified quota summary across all configured providers. |
get_provider_quota |
Detailed quota for one provider (claude, kimi, or zai). Supports an optional mock flag for testing. |
check_quota_health |
Flags providers over a usage threshold (default 80%) or missing/invalid tokens. |
Install
Requirements
- Node.js ≥ 18
- A bearer token for each provider you want to query (see Token setup below)
One-line install
The package is installable directly from GitHub today (no npm account required):
npx -y ryan-knowone/quota-dashboard-mcp
Once the package is published to npm, the canonical command will be:
npx -y quota-dashboard-mcp@latest
Claude Code
Add the server to your Claude Code config (~/.claude/CONFIG.json or via /mcp):
{
"mcpServers": {
"quota-dashboard": {
"command": "npx",
"args": ["-y", "ryan-knowone/quota-dashboard-mcp"],
"env": {
"CLAUDE_TOKEN": "your_claude_oauth_token",
"KIMI_TOKEN": "your_kimi_platform_api_key",
"ZAI_TOKEN": "your_zai_bearer_token"
}
}
}
}
Cursor
Open Cursor Settings → MCP → Add new MCP server, then paste:
- Name:
quota-dashboard - Type:
command - Command:
env CLAUDE_TOKEN=your_claude_oauth_token KIMI_TOKEN=your_kimi_platform_api_key ZAI_TOKEN=your_zai_bearer_token npx -y ryan-knowone/quota-dashboard-mcp
VS Code
Add to your VS Code settings.json (requires the Claude AI extension or any MCP-compatible extension):
{
"mcp": {
"servers": {
"quota-dashboard": {
"command": "npx",
"args": ["-y", "ryan-knowone/quota-dashboard-mcp"],
"env": {
"CLAUDE_TOKEN": "your_claude_oauth_token",
"KIMI_TOKEN": "your_kimi_platform_api_key",
"ZAI_TOKEN": "your_zai_bearer_token"
}
}
}
}
}
Token setup
Claude Code Max
The quota endpoint requires an OAuth token from an authenticated Claude Code browser session. The easiest source is ~/.claude/credentials.json.
The server currently reads CLAUDE_TOKEN from the environment. To extract it:
# macOS
jq -r '.accessToken' ~/Library/Application\ Support/Claude/credentials.json
# Linux
jq -r '.accessToken' ~/.claude/credentials.json
Then set CLAUDE_TOKEN to that value.
Kimi
Use a platform API key from platform.moonshot.cn. Set it as KIMI_TOKEN.
In practice, the same sk-kimi-... key used by Claude Code's Anthropic-compatible proxy also works for the Kimi usage endpoint. If your proxy key returns Invalid Authentication, create a dedicated platform API key from platform.moonshot.cn.
Z.ai
Use a Bearer token from your Z.ai account/dashboard. Set it as ZAI_TOKEN.
Local development
git clone https://github.com/ryan-knowone/quota-dashboard-mcp.git
cd quota-dashboard-mcp
npm install
# Run directly with tsx
CLAUDE_TOKEN=... KIMI_TOKEN=... ZAI_TOKEN=... npm run dev
# Or build and run
npm run build
CLAUDE_TOKEN=... KIMI_TOKEN=... ZAI_TOKEN=... npm start
Testing a tool call
With the server running over stdio, send a JSON-RPC tools/call request:
{
"jsonrpc": "2.0",
"id": 1,
"method": "tools/call",
"params": {
"name": "get_provider_quota",
"arguments": { "provider": "kimi", "mock": true }
}
}
Privacy
- Tokens are read from environment variables at call time.
- Tokens are never written to disk (other than the env vars you already manage).
- Tokens and usage data are never sent to telemetry or any third party except the provider's own API.
License
MIT
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。