Every answer shows its receipt.
You see what a message will use and what it may cost before anything is spent, and what it really cost when it is done.
A free check before anything runs
When you send a message, Cortex66 first checks it on its own servers, at no cost: your permission, the workspace, which models you can use, the sources, whether the estimate fits your cap and whether the price is verified. Your message is saved before any of this, so nothing is lost.
What the receipt shows
The receipt names the model that will answer and why it was chosen, the other options, the tools it may use, the sources it will read, your cap and an estimate range. Cheaper models are tried first.
It also shows the Memory box: the facts it recalled from memories you confirmed, or why it recalled none.
Before and after
Starting the run reserves the cost within your cap; a message whose estimate does not fit the cap does not start. When the answer is done, the receipt shows the estimate, what was reserved and what was settled. "How it was made", under the answer, opens the path the answer took.
When nothing fits
If no model you can use fits the message, the receipt says so and names the one step that fixes it, for example connecting a provider. Nothing is sent and nothing is spent.
No double charges
Sending the same message again, for example after a lost connection, answers with the same result. It does not run or charge a second time.
Related: the cost cap. See everything Cortex66 can do on Capabilities.