Token Usage
Tokens are the unit of measurement that AI models use to process text. Understanding token usage helps you manage your plan budget and get the most out of your workspace.
What Is a Token?
A token is roughly equivalent to four characters of text, or about three-quarters of a word. Both the text you send to your agent and the text your agent sends back are counted in tokens. Context like your system prompt, conversation history, and any tool results also contribute to the token count.
Per-Message Usage Breakdown
Every message in the chat panel shows a usage summary. Click on any message to expand its details:
- Input tokens — the total tokens sent to the model for that request, including your message, the system prompt, conversation history, and any injected context.
- Output tokens — the tokens produced by the model in its reply.
- Cached tokens — tokens that were served from the model's cache rather than reprocessed. Cached tokens are cheaper and faster, so a high cache hit rate is a good sign.
- Total time — the end-to-end latency for that response.
Daily Token Budgets
Your token budget determines how much AI usage you can perform each day. Budget tiers vary by plan:
- Free plan — a limited daily token budget, suitable for light usage and exploration.
- Builder plan — a higher daily budget for regular professional use.
- Founder plan — the highest budget tier, designed for power users and teams with heavy workloads.
Your budget resets automatically every day at midnight UTC.
Tracking Your Usage
You can monitor your current token consumption directly in your workspace. Check the usage indicator in the workspace header or visit your account settings for a detailed breakdown.
For plan details and upgrade options, see Plan Overview. To understand how model choice affects token consumption, see Choosing a Model.