Trial credits
Your 14-day trial includes 3 credits total, shared across your account for agent runs and content generation. Extra projects, trial extensions, and adding a payment method do not replenish this allowance. Unused trial credits expire. During the trial, agents default to an economical available model. You can choose another model and review its estimate before running. Scheduled agents become available when paid billing begins. New runs require enough credits for their reservation; runs already in progress finish under the normal settlement rules. When paid billing begins, your trial remainder is replaced by your full plan allowance, such as 20 credits a month on AI Search Plus. See Plan entitlements. Previously issued trial allowances are preserved until they expire or paid billing begins.Model picker
The Model picker appears when you start or schedule a built-in agent and when you create a custom agent. It shows only models that are available to your account. Fusion is listed first and is the default when it is available; otherwise the default is Claude Opus 5.5. Fusion picks a task-appropriate model for each step, so its band is a range, shown as Best fit (min – max), and retries can raise a run’s cost up to the bound the run dialog shows. Generate draft on an action defaults to Gemini 4 Neon instead. The model catalog assigns these bands to the other models:
A band compares the model’s representative input/output cost with the 1x reference. It is not a fixed charge and does not mean every 2x run costs exactly twice a particular 1x run. Prompt size, output length, tool work, web searches, and cached tokens affect the final amount.
Web searches
Agents can search the public web. Each search is charged at cost: US$0.007, converted to credits at the same rate as model usage, which is about 0.0093 credits per search. Searches count toward the run’s estimate and its final cost. Failed searches aren’t charged.Owned-content reranking
When an agent searches your owned content, the search finds matching passages within your pages and ranks them with a reranking model. While some of your project’s pages are still being indexed, the agent searches whole pages instead, and nothing is ranked or charged. Each ranking is charged at what the provider billed for it, converted to credits at the same rate as model usage. Its price depends on how much text is ranked: typically about US$0.001 per search, or about 0.0013 credits. A ranking that fails isn’t charged, and the search returns its passages in their original search order. Reranking counts toward the run’s estimate and its final cost.Estimated per run
Estimated per run comes from this agent’s recent completed runs, taken from the first of these that has any:- Runs in this project.
- Runs in your account’s other projects. The dialog says Estimated from this agent’s runs in your other projects.
- For built-in agents only, runs across DevTune. The dialog says Estimated from this agent’s runs across DevTune. A custom agent never uses runs from outside your account.
Estimated monthly
The schedule dialog also shows Estimated monthly when a per-run estimate exists:- Daily schedules use 30 runs per display month
- Weekly schedules use
30 ÷ 7runs per display month - Monthly schedules use one run per display month
Balance checks and final cost
A run whose project has no evidence the agent can read is stopped before the reservation, so it never reserves or spends credits. See the evidence a run needs. Before a run starts, DevTune reserves the estimated amount. If no estimate exists, it reserves one credit. The run is blocked when the available account balance cannot cover that reservation. After the run finishes:- The final credit cost is calculated from measured model usage, web searches and owned-content reranking
- Any unused reservation returns to the account balance
- Additional usage is debited from the remaining balance when available
- The Credit transactions ledger replaces the pending estimate with the final amount