/ api-reference / count-tokens

Count Tokens

Estimate input token count for a Messages request.

POST /v1/messages/count_tokens

Anthropic-compatible token counting for context-window planning. Given the same messages you would send to Messages, returns an approximate input token count — useful for deciding whether to truncate before spending a request.

Request

POST/v1/messages/count_tokens

curl https://api.basemnt.ai/v1/messages/count_tokens \
  -H "x-api-key: $SK_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "z-ai/glm-5.2",
    "messages": [
      { "role": "user", "content": "Write a bash function that waits for a port." }
    ],
    "system": "You are a terse coding assistant."
  }'

Body accepts the same model, messages, system, and tools fields as /v1/messages.

Response

200-response.json

{ "input_tokens": 48 }

Approximation

The count is a heuristic estimate, not a tokenizer-exact number. Different models tokenize differently; budget ~5% headroom before assuming a request fits.

Authentication is required — a missing key is a 401, and an invalid key is rejected the same as a model call.