Claude Quota Proxy

anthropic-quota-proxy

A local HTTP proxy that makes Claude Code aware of its own usage limits.

Works with Claude Code (CLI) only. The web chat and browser extension make requests directly from Anthropic's infrastructure — they don't go through a local proxy and can't use this.

Claude Code (Max plan) shows usage bars for the 5-hour and 7-day quota windows in the UI. The model itself has no access to those values — there's no API, no tool, no hook that exposes them during a conversation. This proxy sits between Claude Code and api.anthropic.com, captures the rate-limit headers on every response, and writes a one-line status file that Claude can read any time.

How it works

Claude Code respects the ANTHROPIC_BASE_URL environment variable. Set it to http://127.0.0.1:4080 and all API traffic routes through the proxy. The proxy forwards every request unchanged, reads the rate-limit response headers, and writes ~/.claude/usage-status.md. Claude reads that file when it needs to know the current quota state.

The proxy is a single Node.js script with zero npm dependencies. On Windows it runs as a Windows service via NSSM. On macOS and Linux the same script runs via launchd or systemd.

Output

A single line in ~/.claude/usage-status.md, updated on every inference response:

! means that window has an allowed_warning status — you're close to the limit.

The three pools:

- 5h — rolling 5-hour window
- 7d — rolling 7-day window
- overage — paid burst pool, available once the 7d pool is exhausted

One unified pool covers all models. There is no separate Sonnet or Opus pool despite the Claude Code UI showing individual model bars.

Setup

Windows

Requires Node.js and an admin PowerShell session. NSSM is downloaded automatically.

The script validates Node.js, confirms your…