Auto-Instrument LLM Calls

2Signal provides thin wrappers around popular LLM provider clients. Wrap the client once and every call is automatically traced — no manual span creation required. The wrappers capture the model name, input messages, output text, token usage (prompt / completion / total), and estimated cost.

Python + OpenAI

import twosignal
from openai import OpenAI

client = twosignal.TwoSignal(api_key="your-api-key")
openai_client = twosignal.wrap_openai(OpenAI())

@client.observe(name="my-agent", span_type="AGENT")
def ask(question: str) -> str:
    response = openai_client.chat.completions.create(
        model="gpt-4o",
        messages=[{"role": "user", "content": question}],
    )
    return response.choices[0].message.content

result = ask("What is 2Signal?")
client.flush()

Python + Anthropic

import twosignal
from anthropic import Anthropic

client = twosignal.TwoSignal(api_key="your-api-key")
anthropic_client = twosignal.wrap_anthropic(Anthropic())

@client.observe(name="my-agent", span_type="AGENT")
def ask(question: str) -> str:
    response = anthropic_client.messages.create(
        model="claude-sonnet-4-20250514",
        max_tokens=1024,
        messages=[{"role": "user", "content": question}],
    )
    return response.content[0].text

result = ask("What is 2Signal?")
client.flush()

TypeScript + OpenAI

import { TwoSignal, wrapOpenAI, observe } from "twosignal";
import OpenAI from "openai";

const ts = new TwoSignal({ apiKey: "your-api-key" });
const openai = wrapOpenAI(new OpenAI());

const ask = observe({ name: "my-agent", spanType: "AGENT" }, async (question: string) => {
  const response = await openai.chat.completions.create({
    model: "gpt-4o",
    messages: [{ role: "user", content: question }],
  });
  return response.choices[0].message.content;
});

await ask("What is 2Signal?");
ts.flush();

What Gets Captured

Every wrapped LLM call produces a span with the following fields populated automatically:

Other Supported Providers

The Python SDK ships with wrappers for many additional providers. Each follows the same pattern — wrap the client and all calls are traced automatically.

• Google Gemini
• Mistral
• Cohere
• Groq
• LangChain
• LlamaIndex
• CrewAI
• AutoGen

The TypeScript SDK supports OpenAI, Anthropic, and Vercel AI SDK.

What's Next

Now that your LLM calls are traced, set up evaluators to automatically score every response: Evaluate Agent Outputs.