Show what each bot spends

[aivsomkar]

In plain language

Before: the engines reported tokens (and Claude a dollar cost) on every turn, and the app threw it away. It also turned out the per-driver token reports meant different things (a per-call delta on Claude, a running thread total on Codex), so summing them naively double-counted.

What changes in the app: a small chip in the chat header shows what the current task has spent — e.g. 12.4k tok · $0.06 (dollars only for Claude, the only engine that reports a price). Bot settings get a Usage card; App settings get a Usage tab ranking every bot by cost then tokens with a total — the "which of my bots is costing me money" answer. Cost is captioned honestly: on a subscription it's an equivalent, not a charge. Figures start from zero on upgrade.

---

In plain terms

An app that runs agents unattended should be able to say what that costs. Cost/token data was already on the wire from the drivers and the UI discarded it — but two things the plan assumed turned out to be wrong, and both shaped this:

1. Only Claude reports a dollar cost. Codex, grok, ACP and Antigravity send cost: null. So this is token-first, dollars where known.
2. thread.token-usage.updated means different things per driver — a per-API-call delta on Claude, the thread's running total on Codex, a per-step figure and a turn total on Antigravity. Summing it naively double-counts. The harness never had a consistent figure to keep.

Changes

• Contract (additive): turn.completed.usage?: { input, output } — this turn's total, the one figure the harness banks. Each driver computes it from what it already has: Claude result.usage (input + cache read + cache creation), Codex tokenUsage.last ?? total, Antigravity result.usage. The live indicator is unchanged and never summed. ProviderSnapshot.billing?: "metered" | "subscription" — Claude and Codex both deliberately strip API keys and run on the CLI's own login, so both report subscription; a Claude cost is an equivalent and the UI says so.
• Harness: store.addTaskUsage() banks each settled turn onto TaskRecord.usage { input, output, costUsd, turns } (cost stays null until any turn reports one); flows to clients through the existing wireTask. Bot and fleet totals are client-side sums.
• UI:
  • Chat header chip for the open task — 12.4k tok · $0.06 (cost only when known); hover for turns / in-out split / billing caption; click → bot settings.
  • Bot settings Usage card: this bot across all tasks, captioned "equivalent — on your subscription, not billed" / "billed to your API key" / "doesn't report a price; tokens are counted".
  • App settings Usage tab: bots ranked by cost then tokens, fleet total row — the "which of my bots is costing me money" answer.

Figures start from zero on upgrade (older turns weren't recorded per task). Out of scope: reset/period windows, per-model breakdown, charts.

Item 1.5 of docs/plans/agent-harness-upgrades-v2.md.

Test plan

• Driver tests assert turn.completed.usage on the existing fakes: claude ({12, 5} incl. cache read), codex ({7, 3} from a thread total), antigravity ({105, 20} from result.usage)
• server/store.test.ts (+4): banks and counts turns; cost null until reported then sums only reported; token-less turn still counted; unknown task ignored
• src/lib/usage.test.ts (5): compact tokens, small dollars stay visible, chip text, cross-task sums, billing captions
• pnpm typecheck clean; pnpm vitest run green (65 files, 536 passed)
• Manual in the dev app: Claude turns show the chip with $; bot Usage card and app-settings Usage tab populate; /api/instances reports billing: subscription for claude and codex

🤖 Generated with Claude Code

[coderabbitai]

Warning

Review limit reached

@aivsomkar, you've reached your PR review limit, so we couldn't start this review.

Next review available in: 44 minutes

Limit details: You’ve used all 3 included reviews currently available under your plan.

Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available.
You're only billed for reviews past your plan's rate limits ($0.25/file).

How can I continue?

After more reviews become available, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews.

How do review limits work?

CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability.

For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window.

Please refer docs for additional details.

Review details

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: 2ec2d584-f8ae-438a-afba-9b59301bb811

📥 Commits

Reviewing files that changed from the base of the PR and between ef784ac and 597b744.

📒 Files selected for processing (18)

• server/contracts.ts
• server/drivers/antigravity.test.ts
• server/drivers/antigravity.ts
• server/drivers/claude.test.ts
• server/drivers/claude.ts
• server/drivers/codex.test.ts
• server/drivers/codex.ts
• server/index.ts
• server/store.test.ts
• server/store.ts
• server/testing/fake-claude-cli.ts
• src/components/ChatView.tsx
• src/components/SettingsModal.tsx
• src/components/SettingsPanel.tsx
• src/components/UsageSection.tsx
• src/lib/usage.test.ts
• src/lib/usage.ts
• src/state/store.tsx

---

_{Comment @coderabbitai help to get the list of available commands.}