Docs

Cost and reuse

Flow Companion is free. Your AI provider bills your account for each question, at its own rates.

The first question sends the whole flow

The first question in a chat sends the whole flow along with it. On a large flow, that's usually the most expensive question in the chat. With an Anthropic or Google key, the flow is also sent when it opens in the panel, and when you switch models, to check that it fits the model you picked. Anthropic doesn't charge for that check.

Follow-ups reuse it

Each later question in the same chat sends the flow and the conversation so far. The provider usually reads the flow back from its memory, which is faster and cheaper than the first time. Anthropic keeps it in memory for an hour after each question. OpenAI and Google decide how long they keep it. After a longer break, the next question costs about as much as a first one.

Where to see the numbers

After an answer, hover over the gauge icon next to Send for the provider's numbers on the last question. Click it for three counts: what was sent, what was reused from the provider's memory, and the length of the answer. The counts are in tokens, the unit your provider bills by, and come from its own usage report.

Long chats

When a chat gets long, the panel shows This chat is getting long with a New chat button. The new chat keeps the same flow, and its first question sends the whole flow again. If a flow is too big for the model you picked, the panel says so before you can ask, and no question is sent. It names a model that can read the flow, or another provider to try.

What affects the bill

The size of the flow, the number of questions, the length of the chat, the model you pick, and the Detail setting. Thorough answers take longer and use more of your provider's credit. After you switch models mid-chat, the next question costs about as much as a first one. On Anthropic, changing Detail mid-chat does the same.

Previous: How answers stay grounded · Next: Limits · All docs