[FEATURE]: Show OpenCode Go plan usage in the TUI prompt bar

[Tonlex]

Feature hasn't been suggested before.

• I have verified this feature I'm about to request hasn't been suggested before.

Describe the enhancement you want to request

The TUI already shows tokens (context %) · $cost (session-local) in the prompt bar. Go plan subscribers have no way to see their account-wide plan usage from the CLI — the only place it's visible is the web dashboard, so users find out they're near a limit when the gateway starts throttling.

Proposal: append the Go plan's monthly usage percentage next to the existing cost indicator, e.g.:

123K (62%) · $0.15 · Go 4%

The percentage comes from the official endpoint added in #16513 (GET https://opencode.ai/zen/go/v1/usage, Bearer key auth), so no dashboard scraping or hardcoded limits. Only rendered when an opencode-go key is configured; degrades silently otherwise.

Proposed design

Server (~100 lines, packages/opencode)

• New endpoint that looks up the stored opencode-go API key (existing Auth service), calls /zen/go/v1/usage, returns the monthlyUsage.percent value.
• 60s TTL cache with background refresh: renders return the stale value instantly while the fetch happens async. (60s, not 5min — usage can move fast on expensive models; community Claude statuslines converge on 30s–2min for quota displays.)

TUI (~20 lines, packages/tui/src/component/prompt/index.tsx)

• Extend the existing usage memo (line ~266) and its render (line ~1665): append · Go N% when data is available.
• Between fetches, estimate the delta locally: each completed message on an opencode-go model already carries token usage, and per-model cost rates are known — so show Go ~4.5% (tilde = estimated) that ticks up as you work, snapping to the authoritative value on the next fetch. No extra API calls.

Mockup

┌─────────────────────────────────────────────────────────────┐
│ Ask anything... "refactor this to use Effect"                │
│ Build · opencode-go/deepseek-v4-flash · opencode             │
└─────────────────────────────────────────────────────────────┘
             123K (62%) · $0.15 · Go 4%   ⌘K commands   ⌘I agents

Note: $0.15 is session cost; Go 4% is account-wide (all sessions/machines using the key) — the Go label carries that distinction.

Out of scope for this issue (follow-ups)

• User-configurable refresh interval / visibility toggle
• USD amounts ($2.40/$60) — blocked on exposing limits in the API (see [FEATURE]: Expose balance and monetary usage data via the official usage API #41981)
• Rolling 5h / weekly windows — data is free from the same endpoint; display is a trivial extension
• Gateway-provided rate-limit headers on ordinary responses (Claude Code-style, zero polling) — cheaper long-term than client polling

Related

• feat(console): add go usage endpoint #16513 — merged usage endpoint this builds on
• [FEATURE]: Add Go plan usage/balance API endpoint (rolling/weekly/monthly windows) #16017 — original request for the usage API
• [FEATURE]: zen usage/billing information available in tui #16036 — similar ask for Zen credits in TUI, closed not planned before the endpoint existed; this is scoped to Go plan windows via the official API
• [FEATURE]: Expose balance and monetary usage data via the official usage API #41981 — monetary usage fields (follow-up dependency)

[github-actions]

This issue might be a duplicate of existing issues. Please check:

• [FEATURE]:U、account usage display on the tui sidebar #41293: Requests displaying Go plan usage (rolling/weekly/monthly) in the TUI sidebar — very similar goal of surfacing account-wide usage in the TUI
• [FEATURE]: API endpoint for Go plan usage metrics (rolling/weekly/monthly) #38824: Requests the API endpoint for Go plan usage metrics that this feature builds on — may be worth coordinating
• OpenCode free tier cuts off service mid-task with no warning - expose remaining allowance / add countdown #40588: Requests exposing remaining allowance / countdown in the TUI before quota exhaustion — overlapping problem space

[Tonlex]

Thanks for the flag. This is complementary to #41293 (same data proposed for the TUI sidebar) — this one scopes it down to the minimal prompt-bar variant, and builds on the usage endpoint requested in #38824. Happy to consolidate or adjust the scope if #41293 ends up covering this.