Skip to main content
Get up and running with Cascade in minutes. This guide covers installation, API key setup, and sending your first trace.

​
Install the SDK

Install the Cascade SDK from PyPI: 
pip install cascade-sdk
Requires Python 3.8 or higher.

​
Set your credentials

Get your API key and endpoint from the Cascade Dashboard. Both are provided per customer. Create a .env file in the root of your project:
.env
CASCADE_API_KEY="csk_live_..."
CASCADE_ENDPOINT="https://your-endpoint.runcascade.com/v1/traces"
Or pass them directly in code: 
from cascade import init_tracing

init_tracing(api_key="csk_live_...", endpoint="https://your-endpoint.runcascade.com/v1/traces")

​
Initialize tracing

Call init_tracing() once at the top of your application. Everything else is automatic. 
from cascade import init_tracing

# With a project name (recommended - organizes traces in the dashboard)
init_tracing(project="customer_support_chatbot")

​
Send your first trace

Wrap your agent’s entry point with trace_run() and use wrap_llm_client() to automatically capture all LLM interactions:
my_agent.py
from cascade import init_tracing, trace_run, wrap_llm_client
from anthropic import Anthropic

# Initialize tracing
init_tracing(project="my_first_agent")

# Wrap your LLM client - all calls are now automatically traced
client = wrap_llm_client(Anthropic())

# Create a traced agent run
with trace_run("SimpleAgent"):
    response = client.messages.create(
        model="claude-sonnet-4-20250514",
        max_tokens=100,
        messages=[
            {"role": "user", "content": "Say hello and introduce yourself in one sentence."}
        ]
    )
    print(response.content[0].text)
Run the script: 
python my_agent.py
Open the Cascade Dashboard to see your trace with full details including prompts, completions, token counts, latency, and cost.

​
Next steps

Sessions & Tracing

Learn about tracing runs, sub-agents, tools, and LLM clients.

Evaluation SDK

Score your agent’s performance with built-in and custom evaluators.

Integrations

Auto-instrument LangGraph, OpenAI Agents, and Claude Agents.

Model providers

Trace any LLM provider through a unified wrapper.