Quickstart

Install the Framewren SDK, connect a provider, and see your first cost and latency data in the dashboard. Estimated time: under 10 minutes.

1. Create a free account

Sign up at framewren.com/login/signup.html. No credit card required. The free tier includes 2 provider connections and 50K tracked API calls per month.

After signup, your first project is created automatically. Copy your API key from Settings → API Keys.

2. Install the SDK

# Python (requires 3.8+)
$ pip install framewren-sdk

# Node.js (requires 18+)
$ npm install @framewren/sdk

3. Initialize and wrap your LLM client

Framewren wraps your existing LLM client. You don't change your call sites — just wrap the client at initialization.

import framewren
from openai import OpenAI

# Initialize Framewren with your API key
framewren.init(api_key="frw_live_your_key_here")

# Wrap your existing LLM client
client = framewren.wrap(OpenAI())

# Use client normally — Framewren tracks in the background
response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Hello"}]
)

That's it. The client behaves identically to the unwrapped version — same methods, same response shapes, same error handling. Framewren intercepts at the transport layer.

4. (Optional) Add metadata tags

Tag requests with arbitrary metadata to enable cost attribution by feature, customer tier, or experiment.

response = client.chat.completions.create(
    model="gpt-4o",
    messages=[...],
    extra_headers={
        "X-Framewren-Tags": "feature=summarizer,customer_tier=pro"
    }
)

5. Check your dashboard

Within 30 seconds of your first instrumented call, your dashboard will show:

• Cost per request and cost per hour
• P50, P95, and P99 latency
• Token breakdown: input vs output tokens
• Request volume over time