tokensor 0.1.0

LLM token observability SDK — wraps OpenAI and Anthropic clients to capture metadata

tokensor

LLM token observability for Python. Wraps your OpenAI or Anthropic client to capture call metadata (model, tokens, cost) and forward it to your Tokensor dashboard — one line of code, zero latency overhead.

Install

pip install tokensor
# With OpenAI support:
pip install "tokensor[openai]"
# With Anthropic support:
pip install "tokensor[anthropic]"

Quickstart

Get your API key from app.tokensor.com → Settings → API Keys.

OpenAI:

import os
import openai
import tokensor

client = tokensor.wrap(
    openai.OpenAI(),
    api_key=os.environ["TOKENSOR_API_KEY"],
)

# Use exactly like openai.OpenAI()
response = client.chat.completions.create(
    model="gpt-4o-mini",
    messages=[{"role": "user", "content": "Hello"}],
)

Anthropic:

import os
import anthropic
import tokensor

client = tokensor.wrap(
    anthropic.Anthropic(),
    api_key=os.environ["TOKENSOR_API_KEY"],
)

message = client.messages.create(
    model="claude-3-5-haiku-latest",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Hello"}],
)

Environment variables

Variable	Required	Description
TOKENSOR_API_KEY	yes	Your Tokensor API key
TOKENSOR_HOST	no	Override API host (default: https://api.tokensor.com)

License

MIT