Academy
8. Costs and Best Practices
Understand how mission costs are calculated in credits, read the cost on a mission page, and apply practical tips to keep spending efficient while getting great results.
Every mission you give @oupi has a cost measured in credits. The cost is proportional to the total work done — specifically, the tokens exchanged with AI models across all the rounds of planning, execution, and verification that make up the mission. A short, well-briefed mission ("summarize this PDF in five bullet points") might cost a few dozen credits, while a deep research article drawing on many sources will cost more. The exact credit cost is always displayed on each mission's page after it completes, so you're never guessing. Credits are debited from your balance once the mission finishes.
Two factors drive mission cost: scope and clarity. Scope is how much work @oupi has to do — more rounds, more tool calls, more content produced means more tokens and therefore more credits. Clarity is how well you defined the goal upfront. A vague prompt forces @oupi to explore more possibilities, ask more clarifications, and sometimes produce output you'll want to redo. A precise prompt — stating the outcome, its audience, the format, and any constraints — lets @oupi plan efficiently and reach the result in fewer rounds. In short: clear goals and sensible scope are the best way to keep costs down.
State the outcome and its use, not the steps. Instead of listing ten instructions, tell @oupi what you need, for whom, in what form, and any hard constraints (deadline, sources, things to avoid). It decides the steps itself — and usually picks a shorter path than you would have scripted.
Use follow-ups instead of new missions. When a result is close but needs tweaks, reply from the mission page ("shorten it to one page", "add 2025 figures"). This starts a child mission that already has the parent's context, so @oupi doesn't redo work from scratch — saving both time and credits.
Answer clarifications quickly and concisely. When @oupi asks a question before starting, a brief, direct answer lets it proceed without extra rounds. If you'd rather it decide, just say "use your best judgment" — that's cheaper than leaving a mission waiting and then restarting it.
Always check the verification verdict before using a result. Every mission ends with one of three outcomes: verified (the result matches your goal), failed (it doesn't), or could not be verified (the checker couldn't read the output — e.g., a binary file or an external action). "Could not be verified" is not a failure; it simply means you should review the deliverable yourself. Catching issues at this stage is far more efficient than discovering them later and launching a correction mission.
Open the Assistant → Missions page. Pick a completed mission and look at its credit cost and verification verdict. Then compare it with another mission of a different size. Notice how the cost scales with the complexity of the work. This builds your intuition for estimating future mission costs.
Mission costs are measured in credits, proportional to the tokens used across all rounds. To keep spending efficient: write clear goals with the outcome, audience, and format upfront; keep scope sensible; use follow-ups instead of new missions for tweaks; answer clarifications quickly; and always read the verification verdict before acting on a result. Check costs on any mission page to build your intuition over time.