@tokeneyez/mcp

@tokeneyez/mcp

Bring your TokenEyez AI-spend dashboard into Claude Desktop, Cursor, and any other MCP-compatible client.

TokenEyez tracks your token spend across Claude, ChatGPT, Gemini, and Perplexity. This MCP server lets you query that data — and log new events — straight from inside your AI client.

🔒 Pro feature

The MCP server is available to Pro and Teams subscribers ($9/mo). Free-tier users can sign in and run who_am_i to verify their account, but other tools return a friendly upgrade message. See the full feature comparison and upgrade at https://tokeneyez.tre-lab.com/upgrade.

What you can do

Once installed, ask Claude Desktop / Cursor:

"How much have I spent on AI this month?" "What's my biggest cost driver?" "How many Rephrase analyses do I have left today?" "Log this terminal session — I just used 2,400 tokens on Sonnet."

Claude calls our MCP server, which pulls real numbers from your TokenEyez account.

Install

npm install -g @tokeneyez/mcp

Sign in (once)

Use the same email + password as your TokenEyez Chrome extension:

tokeneyez login

Your session is saved to ~/.tokeneyez/session.json (mode 0600, owner-readable only).

Wire up Claude Desktop

Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

{
  "mcpServers": {
    "tokeneyez": {
      "command": "npx",
      "args": ["-y", "@tokeneyez/mcp"]
    }
  }
}

Restart Claude Desktop. You should see the 🔧 icon appear when the MCP connection is healthy.

Wire up Cursor

Edit ~/.cursor/mcp.json (or use Settings → MCP):

{
  "mcpServers": {
    "tokeneyez": {
      "command": "npx",
      "args": ["-y", "@tokeneyez/mcp"]
    }
  }
}

What the AI can do

Tools (callable by the AI):

Tool	What it does
get_dashboard_summary	Tokens / cost / messages, last N days, broken down by platform
get_savings_summary	Tokens & cost saved by Rephrase + Analyze
get_quota_status	Tier, daily limit, today's usage, top-up credits
record_manual_event	Log a usage event from a tool we don't auto-track yet
who_am_i	Returns the signed-in email + tier

Resources (read-only context):

• tokeneyez://dashboard — last 30 days of usage (JSON)
• tokeneyez://savings — savings summary (JSON)

Prompts (slash commands):

• /tokeneyez:analyze_my_spend [period] — full breakdown + cost-cutting suggestions
• /tokeneyez:show_savings — quick "how much am I saving" reply

CLI commands

tokeneyez login    # sign in (saves JWT)
tokeneyez status   # who am I, tier, today's quota
tokeneyez logout   # clear local session
tokeneyez help     # this help

Privacy

• We read API metadata only (token counts, model names, timestamps).
• No prompt content. No response content. Ever.
• Your session JWT lives only on your machine, in a 0600-permission file.
• See full policy: https://tokeneyez.tre-lab.com/privacy

License

MIT