How Your Credits Are Used in a Conversation
Every time you send a message, the AI needs the whole conversation so far as context. Rather than paying full price for that same context again and again, TeamAI reuses it, and reused context costs less than new text.
This happens in every chat, automatically. The longer a conversation runs, the more of it is being reused, so the first message usually costs the most and the ones after it cost noticeably less.
Knowing this, you can get more out of the same credits.
How to Get More of the Saving
Keep related work in one conversation. Continuing a chat reuses what's already there. Starting a fresh chat and re-pasting the same material means paying full price for it again.
Re-use agents rather than rewriting long prompts. An agent's instructions are part of the reused context, so a team using the same agent repeatedly gets more of the saving than one pasting long prompts by hand.
Don't re-send documents the chat already has. Once a file or brief is in the conversation, the AI still has it. Attaching it again is new input and bills at full price.
Where to See It
Owners can see this on the Usage page. The Served from cache tile shows what share of your input was reused in the period you're viewing, and the Credits & tokens card splits your usage into reused context and new input.
A higher share means more of your team's work is building on existing conversations rather than starting from scratch.
What Doesn't Change
Your plan's included credits and your overage rate stay the same.
Nothing about how you work needs to change.
Your bill won't rise because of this. The same work costs the same or less.
Frequently Asked Questions
Does this apply to every model?
Most models we offer support it, and those are billed at their lower reuse rate automatically. A few don't support caching at all, so work on those bills at the normal rate throughout. If you want the saving, it's worth checking which model your team defaults to.
My Served from cache share looks low. Is something wrong?
Probably not. Short, one-off questions have little to reuse, so there's nothing to serve from cache. The share climbs when people hold longer conversations, work with agents, or keep returning to the same chat.
Does it apply to the API?
Yes. Requests through the public API are billed the same way.
What about my embedded chatbot?
Chatbots don't use credits at all. They run on chatbot credits, which work differently. See How to Purchase and Manage Chatbot Credits.
Do I need to do anything to switch it on?
No. It applies to everyone automatically.