Skip to main content
Glama

meterboard

See what you spend on AI in one place: subscriptions, metered bills, tokens and limits.

meterboard is a small dashboard and a read-only MCP server. It shows what you have spent so far this month, where the month is heading, what a usual month costs and where you could cut. Bills (what a provider charged) and your own token counts are kept apart and never added together.

It is one static page plus two server functions, written with Node built-ins only. There is nothing to install.

Try it

You need Node 20 or newer.

npm start

Then open http://localhost:8787/usage/. With no settings, meterboard shows invented sample data and asks no one to sign in. The sample data is safe to show; nothing about you is in it.

Run the tests:

npm test

Related MCP server: finops-mcp

Use your own data

meterboard reads usage through a data source. Two come with it:

Source

Set METERBOARD_SOURCE to

What it does

Sample data

fixture (the default)

Invented accounts, plans, tokens and limits, dated from today.

Your own API

http

POSTs each query as JSON to METERBOARD_HTTP_URL, with METERBOARD_HTTP_KEY in a header, and shows the answer.

The answer shape is written out in lib/sources/index.mjs. Any API that answers in that shape works: a small server over your billing exports, a workflow tool's webhook or a database view.

To add a different source, write lib/sources/<name>.mjs, register it in lib/sources/index.mjs and make it pass test/source-contract.test.mjs.

Sign-in

Set one mode with METERBOARD_AUTH:

Mode

Who gets in

Use it for

cloudflare-access

People your Cloudflare Access app lets in. meterboard checks the signed token on every request.

A public address.

token

Anyone who types the shared token (METERBOARD_TOKEN) into the page. The page keeps it for that tab only.

Just you, or a small team.

none

Everyone.

The sample data, or a server only you can reach.

Warning: none with real data lets anyone who can reach the server read your usage. meterboard prints a warning when you run it that way.

meterboard fails closed. If you point it at real data and set no mode, or a mode it does not know, it refuses every request with a message saying what to set. In token mode with no token set, nobody gets in.

token mode is a shared secret, not accounts. Anyone who has it sees everything, so make it long and random (openssl rand -hex 32), serve the page over HTTPS only, and change it if it leaks.

MCP server

POST /api/mcp is a read-only MCP server (Streamable HTTP, JSON answers) with two tools:

  • usage_summary: totals from bills and from your own count, the top accounts, agents and models, and every limit at or past 80%, for this month, last month or the last 7 days.

  • usage_query: the source's answer for any date range, with up to 200 rows.

Clients send Authorization: Bearer <METERBOARD_MCP_TOKEN>. With no token set, the server refuses everyone.

Settings

Every setting is listed, with a placeholder, in .env.example. Set them in your host's environment settings, never in the repository. The most common ones:

Name

What it does

METERBOARD_SOURCE

fixture or http.

METERBOARD_AUTH

cloudflare-access, token or none.

METERBOARD_TIME_ZONE

Where "today" and month edges fall, for example Europe/Berlin. Default UTC.

METERBOARD_LIMITS

Labels for your limit names, as JSON. Mark a limit "hosting": true to count its bill readings as hosting spend.

Deploy

meterboard is laid out for Vercel: vercel.json serves public/ and turns api/ into functions. (The standalone tree has not been deployed on its own yet; the same code runs today inside a private deployment.) Any host that serves static files and runs fetch-style handlers (export default { fetch(request) }) works too; scripts/dev.mjs shows the wiring in about 60 lines.

Before you deploy with real data, set a sign-in mode. Check that /api/usage answers 401 when you are signed out.

Layout

  • public/usage/: the page. view.js turns answers into what the page shows and is shared with the server.

  • api/: the two server functions.

  • lib/: settings, sign-in, data sources, the MCP tools.

  • scripts/dev.mjs: the local server.

  • test/: the tests, run with node --test.

License

MIT. See LICENSE.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    Related MCP Connectors

    • Read-only AI/LLM cost tools with current MCP discovery and attributable Registry transport.

    • Veqen brings together work and coding agent costs per person and team, product inference costs per customer, agent spending, protection incidents, invoice checks, chargeback and the month-end close. Ask your assistant to investigate recorded spending, compare model prices, review provider keys and incidents, record bills and credits, or manage budgets, alerts, teams and spending policies. Your account role and plan govern access. Provider-key shut-off and agent-card controls require the appropriate role and provider permissions. Destructive actions require confirmation, and organizations can require a second person to approve a shut-off. SDK budget blocking is not a guaranteed spending cap. Work and coding agents are free for up to 3 people; product inference is free up to $1,000 in monthly AI spend. Connect your Veqen account for private organization data. The separate public model-price server needs no account. Estimates remain separate from provider invoices. Setup and permissions: https://veqen-app.com/docs/mcp

      https://veqen-app.com/mcp
    • Veqen brings together work and coding agent costs per person and team, product inference costs per customer, agent spending, protection incidents, invoice checks, chargeback and the month-end close. Ask your assistant to investigate recorded spending, compare model prices, review provider keys and incidents, record bills and credits, or manage budgets, alerts, teams and spending policies. Your account role and plan govern access. Provider-key shut-off and agent-card controls require the appropriate role and provider permissions. Destructive actions require confirmation, and organizations can require a second person to approve a shut-off. SDK budget blocking is not a guaranteed spending cap. Work and coding agents are free for up to 3 people; product inference is free up to $1,000 in monthly AI spend. Connect your Veqen account for private organization data. The separate public model-price server needs no account. Estimates remain separate from provider invoices. Setup and permissions: https://veqen-app.com/docs/mcp

      https://veqen-app.com/mcp
    • Read-only MCP for AI usage profiles, leaderboards, stats, and docs; no writes or private data.

      https://usagemax.com/mcp

    Related MCP Servers

    • A
      license
      Not graded
      quality
      C
      maintenance
      Enables AI assistants to query multi-cloud cost and usage data through a uniform read-only toolset, including cost summaries, trends, budgets, forecasts, and optimization recommendations.
      MIT
    • A
      license
      Not graded
      quality
      B
      maintenance
      Enables AI agents to query multi-cloud costs, inventory, waste, and commitments across AWS, GCP, Cloudflare, and OVH through read-only MCP tools.
      Apache 2.0
    • A
      license
      Not graded
      quality
      B
      maintenance
      Enables Claude Code sessions to check account rate-limit headroom, local token spend, and live session status, with read-only MCP tools for limits, summaries, token usage, and session listing.
      2
      MIT