How credits work

Credits are the unit you spend when Agent-Q calls an LLM. 1 credit ≈ 1 skill invocation ≈ ~15,000 AI tokens. A full proposal pipeline (background → literature → framework → design) uses ~30–50 credits. Most researchers spend 50–100 credits per month.

What spends credits

• Chat messages — each assistant response is metered after it completes.
• Pipelines — guided multi-skill runs (e.g. "draft full methodology") consume credits per internal skill call.
• Evaluations — RTPI, FES, peer review simulation use Opus on the Pro plan and cost more per invocation.

What does not spend credits

• Reading existing chapters, citations, or the knowledge graph.
• Exporting, sharing, or downloading your data.
• Invites, account changes, billing, settings.

Resets and rollover

Credits reset at the start of each billing period. Bonus credits granted by promo codes reset the same way — they're a boost for the current period, not a balance that accumulates. If you're bumping into the cap regularly, upgrade.

Hitting the limit

You'll see the spend bar fill up in /settings/billing. At 90% it turns red. Once at 100%, chat calls return a 402 quota_exceeded response with a link to upgrade — your projects stay intact, you just can't invoke new LLM work until the next reset or a plan change.

Watch your trend

/settings/billing shows a 14-day spend sparkline and the 7-day breakdown by reason. Use it to catch runaway pipeline loops or notice when a teammate's usage is climbing.