MCP Server

Bring your AI-spend dashboard into your AI assistant.

What it does

The TokenEyez MCP server runs locally as a subprocess of your AI client and speaks the Model Context Protocol. It exposes 5 tools, 2 resources, and 2 slash-prompts so you can ask your assistant questions like:

"What's my AI spend this month?"

"How much have I saved with Rephrase this week?"

"Log a Gemini call: 1200 input, 800 output, gemini-2.5-pro."

"Am I about to hit my daily Claude quota?"

Supported clients

◐

Claude Desktop

macOS, Windows · official MCP support

⌘

Cursor

Native MCP integration in chat

⚡

Cline

VS Code extension with MCP

Any MCP-compatible client that can spawn a stdio subprocess will work — these are just the three we've tested end-to-end.

Install

Install the MCP package

Run this in your terminal — it pulls the latest published build from npm:
```
npm install -g @tokeneyez/mcp
```
Sign in

This stores a session token locally so the MCP server can talk to your TokenEyez account:
```
tokeneyez login
```
You'll be prompted for your TokenEyez email + password (the same ones you use for the extension).
Wire it up to your AI client

Claude Desktop — edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):
```
{
  "mcpServers": {
    "tokeneyez": {
      "command": "tokeneyez-mcp"
    }
  }
}
```
Cursor — Settings → MCP → Add new server, point at tokeneyez-mcp.

Cline — open the MCP Servers panel and add tokeneyez-mcp as a stdio server.
Restart your client

Restart Claude Desktop (or your IDE) so it picks up the new server. You'll see TokenEyez tools appear in the chat's tool list.
Try it

Ask your assistant something like "What's my AI spend this month?" — it should call get_dashboard_summary and reply with your real numbers.

Available tools

get_dashboard_summary({ days?: number })

Today's totals, range totals, and per-platform breakdown for the last N days (default 30).

get_savings_summary({ days?: number })

Tokens saved, cost saved, and acceptance rate from Rephrase + Analyze.

get_quota_status()

Current tier, daily limit, today's usage, remaining quota, and any top-up credits.

record_manual_event({ platform, model, input_tokens, output_tokens })

Log usage from a tool TokenEyez doesn't auto-track (a CLI, an API call, a desktop app). Appears in your dashboard immediately.

who_am_i()

Returns the email + tier of the signed-in user. Available on the free tier so you can verify your account state.

Resources & prompts

The server also exposes two read-only JSON resources — tokeneyez://dashboard and tokeneyez://savings — and two slash-prompt templates: analyze_my_spend and show_savings. Most clients surface these as autocomplete options once the server is connected.

Privacy

The MCP server reads only the same numeric metadata the extension already records — token counts, model names, cost estimates, timestamps. No prompt or response content ever leaves your device through MCP. See the full Privacy Policy.

Troubleshooting

"Not signed in" in the chat

Run tokeneyez login again. The session refreshes itself, but if you cleared local data the server will need fresh credentials.

"TokenEyez MCP requires a Pro subscription"

Your current tier is below Pro. The MCP server intentionally returns this message instead of failing silently. Upgrade here and the next tool call will work — no restart needed (the server re-checks your tier every five minutes).

The server starts but no tools appear

Most likely the client config points at a binary it can't find. Run which tokeneyez-mcp (macOS/Linux) or where tokeneyez-mcp (Windows) and use the absolute path in the config's "command" field.

Still stuck?

Email tokeneyezofficial@gmail.com with "MCP" in the subject and we'll respond within 1 business day.