OpenCode TUI plugin that shows Codex, OpenCode GO, ZAI, Synthetic, MiniMax Token Plan, and Qwen usage limits in the sidebar and prompt footer.
- Adds a
Usage Limitsblock under the sidebarContextsection. - 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
qwencloudCLI. - 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, or MiniMax 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.
Standard OpenCode uses the stable package from npm's latest dist-tag. Install it globally with:
opencode plugin oc-usage-limits-plugin -g-g/--globalinstalls to~/.config/opencode/tui.json.- Without
-g, installs locally to<project>/.opencode/tui.json(requires a git worktree). --forcereplaces an existing pinned version.
The CLI installs the latest package and updates the TUI plugin config for you. The package entrypoint is oc-usage-limits-plugin/tui.
OpenCode v2 uses the preview package from npm's next dist-tag. Install it globally with OpenCode v2 only:
opencode plugin oc-usage-limits-plugin@next -gThe v2 package is built and released from the opencode-v2 branch. It is preview/beta software until validation against the v2 host is complete. Do not use @next with standard OpenCode, and do not use the stable package with OpenCode v2.
To configure it manually instead, add the plugin to ~/.config/opencode/tui.json:
OpenCode TUI plugins are configured in tui.json, not opencode.jsonc.
Restart OpenCode after changing TUI plugin config.
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 tui.json entry still resolves to an older version.
For standard OpenCode (@latest):
PowerShell:
Remove-Item -LiteralPath "$HOME\.cache\opencode\packages\oc-usage-limits-plugin@latest" -Recurse -ForcemacOS/Linux:
rm -rf ~/.cache/opencode/packages/oc-usage-limits-plugin@latestFor OpenCode v2 (@next):
PowerShell:
Remove-Item -LiteralPath "$HOME\.cache\opencode\packages\oc-usage-limits-plugin@next" -Recurse -ForcemacOS/Linux:
rm -rf ~/.cache/opencode/packages/oc-usage-limits-plugin@nextStart OpenCode again and it will reinstall the plugin from the existing tui.json entry. If the plugin is no longer configured, run the install command again:
opencode plugin oc-usage-limits-plugin -g- Dependency conflicts involving
@opencode-ai/pluginusually 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 availableright 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.
- Stable releases come from
main, use thelatestnpm dist-tag, and are intended for standard OpenCode. - OpenCode v2 previews come from
opencode-v2, use thenextnpm dist-tag, and are intended only for OpenCode v2. - Both lanes use the same package name and
oc-usage-limits-plugin/tuientrypoint. - Releases are staged on npm for manual approval before publication. Stable releases create normal GitHub releases; v2 previews create GitHub prerelease tags/releases.
- The v2 lane remains preview/beta until the package has been validated against the v2 host.
Create ~/.config/opencode/usage-limits.jsonc. The same file lives at examples/usage-limits.jsonc and can be copied verbatim:
{
"$schema": "https://raw.githubusercontent.com/mynameistito/oc-usage-limits-plugin/main/usage-limits.schema.json",
"enabled": true,
"refreshIntervalSeconds": 60,
"requestTimeoutMs": 10000,
"showErrors": true,
"providers": {
"codex": {
"enabled": true,
"label": "Codex",
"showSidebarBar": true,
"showFooterBar": true,
"sidebarWindow": "all",
"footerWindow": "auto",
},
"zai": {
"enabled": true,
"label": "ZAI",
"apiKey": "{env:OC_ZAI_API_KEY}", // Optional fallback when OpenCode auth has no ZAI key
"authorizationScheme": "raw",
},
"synthetic": {
"enabled": true,
"label": "Synthetic",
"apiKey": "{env:OC_SYNTHETIC_API_KEY}", // Optional fallback when OpenCode auth has no Synthetic key
},
"minimax": {
"enabled": true,
"label": "MiniMax",
"apiKey": "{env:OC_MINIMAX_TOKEN_PLAN_KEY}", // Optional fallback when OpenCode auth has no MiniMax key
},
},
}If you only need Codex and ZAI with auto-discovered credentials:
{
"$schema": "https://raw.githubusercontent.com/mynameistito/oc-usage-limits-plugin/main/usage-limits.schema.json",
"providers": {
"codex": { "enabled": true },
"zai": { "enabled": true, "authorizationScheme": "raw" },
},
}Disabled providers are hidden:
"providers": {
"codex": { "enabled": true },
"zai": { "enabled": false }
}enabled is the plugin master switch, while each provider's enabled field controls fetching. Provider showSidebarBar and showFooterBar independently control that provider's displays without stopping refreshes. sidebarWindow filters that provider's sidebar windows (all, rolling, daily, weekly, monthly, credits, or other), and footerWindow selects its footer window (auto or one of those kinds). Both display flags default to true; sidebarWindow defaults to all and footerWindow to auto.
| 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 | — |
opencode-go |
OpenCode GO usage | OPENCODE_API_KEY |
Bearer | https://opencode.ai/zen/go/v1 |
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.
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, provideropenai. - Codex auth file from
authPath, default~/.codex/auth.json.
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, providerzai-coding-plan. - OpenCode auth provider
zai. - Config
apiKey, including{env:OC_ZAI_API_KEY}references.
Synthetic lookup order:
- Config
authPathJSON file ({ "key": "..." }/{ "apiKey": "..." }/{ "synthetic": { "key": "..." } }). - OpenCode auth at
~/.local/share/opencode/auth.json, providersynthetic. - Config
apiKey, including{env:OC_SYNTHETIC_API_KEY}references.
MiniMax Token Plan lookup order:
- Config
authPathJSON file ({ "key": "..." }/{ "apiKey": "..." }/{ "minimax-coding-plan": { "key": "..." } }). - OpenCode auth at
~/.local/share/opencode/auth.json, providerminimax-coding-plan,minimax, orminimax-token-plan. - Config
apiKey, including{env:OC_MINIMAX_TOKEN_PLAN_KEY}references.
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 56mPrompt footer shows compact usage when the current session model belongs to a supported provider:
5h: 42% used resets 1h 2mProvider 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);minimaxis also accepted as an alias. - OpenCode provider
qwen-> Qwen Token Plan usage. - OpenCode provider
opencode-go-> OpenCode GO usage.
bun install
bun run typecheck
bun run test
bun run check
bun run buildThe package exposes a TUI entrypoint at oc-usage-limits-plugin/tui for OpenCode's package plugin loader.
- 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-modelcurrent_interval_statusandcurrent_weekly_statusare treated as1= in plan and3= not in plan, and a window is hidden when its status is3(the API otherwise reports a meaningless100%remaining for a non-existent bucket).
{ "$schema": "https://opencode.ai/tui.json", "plugin": ["oc-usage-limits-plugin"], }