Skip to content

About

Z.ai GLM Coding Plan quota & usage for OpenCode — token-efficient tool plus a live status-line badge

Resources

Stars

0 stars

Watchers

0 watching

Forks

Latest commit

 

History

6 Commits

Folders and files

Repository files navigation

opencode-zai-usage

Z.ai GLM Coding Plan quota & usage for OpenCode v2 — one token-efficient tool plus a live status-line badge. No installer, no config surgery, no extra agent round-trips, zero runtime dependencies.

What you get

  • zai_usage tool — compact, accurate quota report for the agent to call on demand: 5-hour credit window, weekly limit, and monthly MCP tool usage, each with percentage, progress bar, and reset time, plus derived absolute credits for your plan tier. detail: true adds the 24-hour model/tool breakdown and a pricing reference block.
  • Status-line badge — ZAI 21% · 2h 41m in the prompt footer, always current (refreshed as usage events land), color-coded at 60% and 90%, with a · peak marker during peak hours. Zero model calls.
  • /zai command — instant dialog with the full quota table, credits, and pricing (your selected model highlighted). Zero model calls.

Credits & pricing

The monitoring API reports percentages of your credit allowance (docs.z.ai/devpack/overview), not raw tokens. The plugin derives absolute credits from the plan tier the API reports (Lite 2,000/10,000, Pro 12,000/60,000, Max 28,000/140,000 per 5h/week) and shows reference pricing:

  • Peak hours are Mon–Fri 14:00–18:00 SGT; everything else is off-peak at 0.5× credits. The badge flags peak windows and the dialog counts down to the next transition.
  • Model credit multipliers (GLM-5.3: 6.9/1.7/24, Flash: 2.3/0.56/8 for input/cached/output) and the MCP per-call cost (1.2) come from the docs, including the documented routing of older model names.

These constants live in src/pricing.ts with their source and check date — quota percentages always come live from the API, and static values are never mixed into live numbers. The China plan (open.bigmodel.cn) has different, undocumented terms, so pricing is global-plan only.

Install

opencode plugin add opencode-zai-usage

One command: installs the package and registers it in your global OpenCode configuration. Updates and removal are managed the same way (opencode plugin update, opencode plugin remove, opencode plugin list).

Prefer manual configuration? Reference the package (or a local checkout) in ~/.config/opencode/opencode.json:

{
  "$schema": "https://opencode.ai/config.json",
  "plugins": ["opencode-zai-usage"] // or: "/absolute/path/to/opencode-zai-usage"
}

Authenticate as usual with /connect → Z.AI Coding Plan. That's the whole setup.

How it works

Z.ai monitoring API            OpenCode server                TUI
────────────────────           ─────────────────              ─────────────────
 /api/monitor/usage/  ──fetch──▶ zai-usage plugin   ──RPC──▶  badge + /zai
 quota/limit               (credentials · TTL cache ·          dialog
 model-usage *             idle-refresh)
 tool-usage  *
  • Credentials come from OpenCode's integration layer (ctx.integration.connection), never from parsing files. Falls back to ZHIPU_API_KEY (or legacy ZAI_API_KEY) when no connection exists.
  • One fetch, many consumers: a shared TTL cache (60s default) backed by in-flight deduplication serves the tool, badge, and dialog. The cache also refreshes on session.usage.updated — every model call of every session, including parallel ones — with a 5-minute drift floor for usage that happens outside this instance, all capped at one fetch per 30s.
  • The API key never leaves the server plugin. The TUI talks to it over the plugin RPC channel and receives parsed snapshots only.
  • Only percentages are reported for token windows — the monitoring API does not expose absolute plan limits, so none are invented.

Platforms

Both Z.ai global (api.z.ai, integrations zai-coding-plan/zai) and China (open.bigmodel.cn, integrations zhipuai-coding-plan/zhipuai) are supported and detected from whichever integration you connected. For env-var users on the China endpoint, set the platform:

{
  "plugins": [{ "package": "opencode-zai-usage", "options": { "platform": "cn" } }]
}

Development

npm install
npm run check        # typecheck
npm run smoke        # live API smoke test (prints parsed quota, never the key)

scripts/smoke.ts reads credentials the same way the plugin does, minus the integration layer.

Local loading: directory plugins need entry files at the package root, so index.ts and tui.tsx re-export from src/ (they are excluded from the npm package via files, where exports takes over). Point plugins at this directory, restart the service, and check with opencode plugin list or opencode api get /api/plugin.

License

MIT

About

Z.ai GLM Coding Plan quota & usage for OpenCode — token-efficient tool plus a live status-line badge

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages