Credits & capacity

WikiFix is billed through Atlassian at a per-user price, and your subscription converts into scan credits — the unit the scanner actually spends. This page explains the conversion, what consumes credits, and what happens when the balance hits zero.

Credits pay for the checks a scan runs. There are two kinds, and the in-product credits screen shows the live rate for each:

• AI checks. Every time the scanner spots related passages, an AI model compares them to decide whether they genuinely contradict. That comparison spends credits, metered by the amount of text analyzed — the more overlapping content your spaces have, the more comparisons a scan runs.
• Non-AI checks. Checks that don’t need an AI model — today, spotting pages whose owner has left your site — charge a small flat rate per finding they raise. No findings, no charge.

Everything else is free: browsing reports, applying fixes, reverting, notifying owners, the page badge. Credits pay for the scan, not for using the results.

Exact per-check credit rates are being finalized ahead of the Marketplace launch and will be published here. The in-product credits screen always shows the current rates.

The allowance rule is one line: 100 credits per dollar of list plan price. Per seat, that means:

Example: a 100-seat site pays $300/month and gets 30,000 credits/month. The allowance is sized to cover a normal scan cadence — weekly scans of your active spaces — with headroom.

Credits arrive with each Atlassian billing transaction and are spendable through the billing period they were bought for: monthly subscriptions get a fresh grant each month, annual subscriptions get the year’s allowance up front.

If you’re on a discounted plan, credits are granted on the full list price — the discount lowers your bill, not your capacity.

The Atlassian trial includes a starter allotment of scan credits — enough to scan a typical space and see real findings — and lasts for your evaluation period. To scan more during the trial, add your own Anthropic API key (next section); the AI analysis then runs on your Anthropic account instead of the trial credits. Non-AI checks still spend credits at their per-finding rate.

The running scan stops cleanly and the run is marked credit exhausted. Findings already produced stay in the report — nothing is lost, and nothing is charged beyond your balance. Scanning picks up again when:

• your next billing period’s grant arrives, or
• you add your own Anthropic API key — then scans don’t stop at all.

If your scan volume outgrows the credit allowance, add your own Anthropic API key in WikiFix Admin → LLM. The AI analysis then runs on your Anthropic account instead of consuming credits, with no volume cap from us on the AI side. Non-AI checks still spend credits at their per-finding rate. You choose the order: key-only, or credits first with the key as an automatic fallback — when the credits run out mid-scan, the key takes over without interrupting the run.

One thing to know before you add a key: with your own key, the AI step is sent to Anthropic’s API under your account and your agreement with Anthropic — not through the EU AWS Bedrock route described on the security page, which covers the default no-key setup. The formal security write-up of the bring-your-own-key path is part of our pre-launch data audit and will land on the security page.

The credits bar at the top of WikiFix shows this month’s remaining grant, any carry-over from earlier grants, and — if you’ve added a key — whether it’s on standby or in use. During a scan the bar updates live as the scan spends credits.