OpenCode TUI plugin that shows Codex, OpenCode GO, ZAI, Synthetic, MiniMax Token Plan, Qwen, and Alibaba Token Plan usage limits in the sidebar and prompt footer.

Features

Adds a Usage Limits block under the sidebar Context section.
Shows current Codex usage windows from OpenAI/Codex auth.
Shows current ZAI quota windows from ZAI Coding Plan auth.
Shows current Synthetic rolling 5-hour and weekly windows.
Shows current MiniMax Token Plan rolling 5-hour and weekly windows.
Shows current Qwen Token Plan windows from the local qwencloud CLI.
Shows current Alibaba Token Plan 5-hour and weekly windows from the local bl CLI.
Shows current OpenCode GO rolling, weekly, and monthly windows.
Adds compact prompt-footer usage when the current session uses an OpenAI, OpenCode GO, ZAI Coding Plan, Synthetic, MiniMax Token Plan, or Qwen Token Plan model.
Providers are toggled from ~/.config/opencode/usage-limits.jsonc.
Reads OpenCode-connected credentials first, then falls back to explicit config/env credentials.

Install

This package contains the OpenCode v2 TUI plugin and is published from the monorepo main branch using npm's latest dist-tag. Install it globally with:

opencode2 plugin add @mynameistito/opencode-usage-limits@latest -g

-g / --global installs to ~/.config/opencode/cli.json.
Without -g, installs locally to <project>/.opencode/cli.json (requires a git worktree).
--force replaces an existing pinned version.

The CLI installs the latest package and updates the TUI plugin config for you. The package entrypoint is @mynameistito/opencode-usage-limits/tui.

To configure it manually instead, add the plugin to ~/.config/opencode/cli.json:

{
  "$schema": "https://opencode.ai/v2/cli.json",
  "plugins": ["@mynameistito/opencode-usage-limits"],
}

OpenCode CLI/TUI plugins are configured in cli.json.

Restart OpenCode after changing TUI plugin config.

Published package contents

@mynameistito/opencode-usage-limits/tui is the TUI plugin entrypoint.
@mynameistito/opencode-usage-limits/schema is the JSON schema for the usage-limits configuration.
examples/usage-limits.jsonc is a complete, annotated configuration example.

Troubleshooting

Reinstall or refresh the cached plugin

If the plugin is stale, broken, or needs a clean reinstall, quit OpenCode and remove the cache for the lane you installed. OpenCode caches immutable package versions, so clearing the cache is required when a cli.json entry still resolves to an older version.

For standard OpenCode (@latest):

PowerShell:

Remove-Item -LiteralPath "$HOME\.cache\opencode\packages\@mynameistito\opencode-usage-limits@latest" -Recurse -Force

macOS/Linux:

rm -rf ~/.cache/opencode/packages/@mynameistito/opencode-usage-limits@latest

Start OpenCode again and it will reinstall the plugin from the existing cli.json entry. If the plugin is no longer configured, run the install command again:

opencode2 plugin add @mynameistito/opencode-usage-limits -g

Dependency conflicts involving @opencode/plugin usually mean OpenCode's package cache contains an older plugin API package. Update OpenCode, clear the cached plugin as above, then retry the install. This package does not publish OpenCode runtime packages as peer dependencies.
No versions available right after a release means a supply-chain cooldown policy (e.g. min-release-age) is blocking the fresh version. Wait for the cooldown window to pass, or install a previously vetted version instead.

Releases

Both packages release from the monorepo main branch using normal root Changesets and npm's latest dist-tag. The package entrypoint is @mynameistito/opencode-usage-limits/tui.

Usage Config

Create ~/.config/opencode/usage-limits.jsonc from the complete examples/usage-limits.jsonc example. It includes every top-level, provider-common, and provider-specific option.

Minimal config

If you only need Codex and ZAI with auto-discovered credentials:

{
  "$schema": "https://raw.githubusercontent.com/mynameistito/opencode-plugins/main/packages/opencode-usage-limits/usage-limits.schema.json",
  "providers": {
    "codex": { "enabled": true },
    "zai": { "enabled": true, "authorizationScheme": "raw" },
  },
}

Disabled providers are hidden:

"providers": {
  "codex": { "enabled": true },
  "zai": { "enabled": false }
}

Top-level enabled is the plugin master switch, and showErrors controls error text globally. Each provider's enabled controls fetching. Provider showSidebarBar and showFooterBar independently control its sidebar and footer displays without stopping refreshes; both default to true.

Each provider's sidebarWindow can be all, rolling, daily, weekly, monthly, credits, or other. Rolling includes legacy 5h labels. Each provider accepts footerWindow with auto (the provider's normal selection), or one of the same window kinds. An unavailable requested footer window falls back to the provider's automatic selection and then its first available window.

Providers

Alibaba Token Plan (Personal/Solo)

Enable alibaba-token-plan to read the real 5-hour and weekly quota windows through the official Bailian CLI. This is separate from the existing qwen / qwencloud integration and from the older Alibaba Coding Plan. Team credit balances are not supported by this adapter.

Install Bailian CLI following its upstream instructions, then authenticate once:

bl auth login --console --console-site international
bl usage token-plan --console-region ap-southeast-1 --console-site international --output json

The CLI owns authentication; the plugin neither imports browser cookies nor stores credentials. Initial console login may open a browser, but refreshes use the API through bl. An OpenCode model API key alone is insufficient. bl must be on the PATH inherited by OpenCode. Command failures (including a missing CLI or expired login) use the plugin's normal error/stale-data display.

Add to ~/.config/opencode/usage-limits.jsonc:

{
  "providers": {
    "alibaba-token-plan": {
      "enabled": true,
      "region": "international",
      "showSidebarBar": true,
      "showFooterBar": true,
    },
  },
}

For mainland China, use region: "china" and log in with bl auth login --console --console-site domestic. The adapter selects cn-beijing / domestic; the default is ap-southeast-1 / international. Footer aliases are alibaba, alibaba-cn, and alibaba-token-plan. Missing quota windows are omitted rather than displayed as zero or unlimited; an entirely empty response is treated as unavailable, preserving last-good data.

The response contract follows the official CLI's usage/token-plan.ts: per5HourPercentage and per1WeekPercentage are fractions from 0 to 1, and reset timestamps are Unix milliseconds.

Provider ID	Service	Env var	Auth header	Default base URL
`codex`	ChatGPT Codex usage	—	Bearer	`https://chatgpt.com/backend-api`
`zai`	Z.AI Coding Plan quota	`OC_ZAI_API_KEY`	raw / Bearer	`https://api.z.ai`
`synthetic`	Synthetic quotas	`OC_SYNTHETIC_API_KEY`	Bearer	`https://api.synthetic.new`
`minimax`	MiniMax Token Plan	`OC_MINIMAX_TOKEN_PLAN_KEY`	Bearer	`https://www.minimax.io`
`qwen`	Qwen Token Plan	`qwencloud` CLI	CLI	—
`alibaba-token-plan`	Alibaba Token Plan	—	CLI	—
`opencode-go`	OpenCode GO usage	`OPENCODE_API_KEY`	Bearer	`https://opencode.ai/zen/go/v1`

Qwen usage requires the local qwencloud CLI to be installed and authenticated because the plugin calls its authentication-status and usage commands; an unauthenticated CLI state appears as missing credentials.

Synthetic always uses Bearer auth and ignores authorizationScheme.

Set baseUrl on minimax to https://api.minimaxi.com when using the mainland-China region. MiniMax always uses Bearer auth and ignores authorizationScheme.

Credential Lookup

authPath and apiKey are optional overrides. Typical OpenCode users only need enabled: true; label is an optional display override. Credentials are discovered automatically from OpenCode auth and provider defaults. Set apiKey (or authPath to a standalone key file) only when auto-discovery is not enough.

Codex lookup order:

OpenCode auth at ~/.local/share/opencode/auth.json, provider openai.
Codex auth file from authPath, default ~/.codex/auth.json.

If an official-host request rejects the OpenCode credentials, the plugin retries with the configured or default Codex auth file.

ZAI lookup order:

Config authPath, which can point at OpenCode auth JSON or a simple { "key": "..." } / { "apiKey": "..." } JSON file.
OpenCode auth at ~/.local/share/opencode/auth.json, provider zai-coding-plan.
OpenCode auth provider zai.
Config apiKey, including {env:OC_ZAI_API_KEY} references.

Synthetic lookup order:

Config authPath JSON file ({ "key": "..." } / { "apiKey": "..." } / { "synthetic": { "key": "..." } }).
OpenCode auth at ~/.local/share/opencode/auth.json, provider synthetic.
Config apiKey, including {env:OC_SYNTHETIC_API_KEY} references.

MiniMax Token Plan lookup order:

Config authPath JSON file ({ "key": "..." } / { "apiKey": "..." } / { "minimax-coding-plan": { "key": "..." } }).
OpenCode auth at ~/.local/share/opencode/auth.json, provider minimax-coding-plan, minimax, or minimax-token-plan.
Config apiKey, including {env:OC_MINIMAX_TOKEN_PLAN_KEY} references.

Display

Sidebar rows look like:

Usage Limits
codex
  5h: 42% used resets 1h 2m
  weekly: 12% used resets 3d 4h
ZAI
  tokens: 18% used resets 2h
  MCP: 6% used
Synthetic
  5h: 0% used resets 11m
  weekly: 11% used resets 7m
MiniMax
  5h: 10% used resets 2h 56m

Prompt footer shows compact usage when the current session model belongs to a supported provider:

5h: 42% used resets 1h 2m

Provider mapping:

OpenCode provider openai -> Codex usage.
OpenCode provider zai-coding-plan -> ZAI token usage.
OpenCode provider synthetic -> Synthetic usage.
OpenCode provider minimax-coding-plan -> MiniMax Token Plan usage (prompt footer); minimax is also accepted as an alias.
OpenCode provider qwen -> Qwen Token Plan usage.
OpenCode provider opencode-go -> OpenCode GO usage.

Development

bun install
bun run typecheck
bun run test
bun run check
bun run build

The package exposes a TUI entrypoint at @mynameistito/opencode-usage-limits/tui for OpenCode's package plugin loader.

Notes

The refresh interval defaults to 60 seconds.
The effective minimum refresh interval is 15 seconds.
Provider work starts in a scoped coordinator after both TUI slots are registered. Disposal interrupts the coordinator and its active provider work.
Errors are intentionally short and do not include auth tokens or response bodies.
MiniMax Token Plan returns { model_remains, base_resp }; the per-model current_interval_status and current_weekly_status are treated as 1 = in plan and 3 = not in plan, and a window is hidden when its status is 3 (the API otherwise reports a meaningless 100% remaining for a non-existent bucket).