Skip to main content
Sign inSign up
Documentation

Credits

What credits are for

Credits meter the panel's AI chat. Every message spends the model's real token consumption, converted to credits — so a short question costs a fraction of what a long, multi-step research task does.

What does not consume credits:

  • Calls from external assistants (Claude, ChatGPT, Cursor…). Their provider bills you for the model; MoodleMCP only brokers the tool calls.
  • Anything else in the panel: managing keys, viewing usage, the Moodle connection.

Packs

Credits are bought as one-off packs — Starter (1,000), Standard (5,000) and Pro (15,000). Larger packs carry a better rate; current prices are on the pricing page. Buy them from Organization → Credits.

Packs don't expire and don't renew. They stack: buying a second pack adds to the balance you already have.

New organizations may receive a welcome grant of credits so you can try the chat before buying anything.

How a message is charged

  1. Before the message runs, the panel holds an estimate based on the maximum tokens that turn could produce.
  2. When it finishes, the hold is settled against actual usage and the difference is released.

So an interrupted or short answer doesn't cost the estimate — only what it really used. If the balance can't cover the estimate, the message is refused before anything runs, rather than failing halfway.

Watching the balance

Organization → Credits shows the current balance and the transaction ledger — grants, purchases and consumption, each with its date. It's the place to look when consumption seems higher than expected: a long agentic conversation, one that re-reads large course structures, spends noticeably more than a quick lookup.

Keeping consumption down

  • Ask narrow questions: one course, one cohort, one date range.
  • Start a new conversation for a new topic — a long history is re-sent with every message.
  • Use external assistants for heavy exploratory work if you're already paying for that model elsewhere.

Each message is capped at 8 tool round-trips and 4096 output tokens, which bounds what a single turn can ever spend.