Token counting via API

Count before you send

Before you send a request, you can count exactly how many tokens it will use via POST /v1/messages/count_tokens. This lets you validate inputs, enforce limits, and predict costs before they happen.

Pass the same payload you'd send to /v1/messages and get back the input token count. No output is generated — it's purely a count.

Where this is useful

• Warn users before they submit a request that's too large
• Dynamically trim context to stay under a budget
• Log token usage per user for billing or rate limiting