Quickstart

Welcome to Trunkate AI. This guide will get you from zero to your first prompt reduction in under 5 minutes.

Prerequisites

Before we begin, ensure you have the following:

A GitHub or Google account (for Dashboard access).
Node.js v18+ or Python 3.9+ installed.

Get your API Key

All interactions with the Trunkate AI API require a secure API key.

Log in to the Trunkate AI Dashboard .
Navigate to Settings > API Keys.
Click “Generate New Key”.
Copy the key immediately (it is only shown once).

WARNING

Your API key carries your financial identity. Never commit it to a public repository. Use .env files.

Install the SDK

We provide native SDK bindings and a global command-line tool.

Typescript


npm install @trunkate-ai/sdk

Python


pip install trunkate-ai

Go


go get trunkate.ai/go

Rust


cargo add trunkate-ai

CLI


npm install -g @trunkate-ai/cli

The “Hello World” of Prompt Optimization

Let’s optimize a Prompt. Instead of sending verbose instructions to the LLM, we will use Trunkate AI to shrink it down.

TypeScript


import { Trunkate } from "@trunkate-ai/sdk";
 
// Initialize client using your API key
const client = new Trunkate({ 
  apiKey: "tk_live_1234567890abcdef" 
});
 
const prompt = "Your very long context...";
 
// Option A: Fixed token budget (integer)
const { text } = await client.optimize(prompt, "Summarize", 500);
 
// Option B: Proactive percentage budget (string)
const { text: optimized } = await client.optimize(prompt, "Summarize", "20%");
 
console.log(`Reduced token count by ${stats.reduction_percent}%`);

Python


import os
from trunkate_ai import Trunkate
 
# Initialize client using your API key
client = Trunkate(api_key="tk_live_1234567890abcdef")
 
prompt = "Your very long context..."
 
# Option A: Fixed token budget
optimized = client.optimize(text=prompt, budget=500)
 
# Option B: Proactive percentage budget
optimized = client.optimize(text=prompt, budget="20%")
 
print(f"Reduced token count by {optimized.stats.reduction_percent}%")

Go


package main
 
import (
    "fmt"
    "trunkate.ai/go"
)
 
func main() {
    // Initialize client using your API key
    client, _ := trunkate_ai.NewClient("tk_live_1234567890abcdef")
    
    // Option A: Fixed
    resp, _ := client.Optimize(&trunkate_ai.Request{
        Text:   "...",
        Budget: 500,
    })
    
    // Option B: Percentage
    resp2, _ := client.Optimize(&trunkate_ai.Request{
        Text:   "...",
        Budget: "20%",
    })
    
    fmt.Printf("Optimized: %s", resp.Text)
}

Rust


use trunkate_ai::Client;
 
#[tokio::main]
async fn main() {
    // Initialize client using your API key
    let client = Client::new(Some("tk_live_1234567890abcdef".to_string()));
    
    let text = "Your very long context...";
    
    // Option A: Fixed
    let res = client.optimize(text, None, Some(trunkate_ai::StringOrInt::Int(500)), None).await?;
    
    // Option B: Percentage
    let res2 = client.optimize(text, None, Some(trunkate_ai::StringOrInt::String("20%".to_string())), None).await?;
    
    println!("Optimized: {}", res.text);
}

CLI


# Install globally
npm install -g @trunkate-ai/cli
 
# Export your API key in your terminal
export TRUNKATE_API_KEY="tk_live_1234567890abcdef"
 
# Fixed budget
trunkate-ai --text "..." --budget 500
 
# Percentage budget
trunkate-ai --text "..." --budget "20%"
# Output stats to stderr

OpenClaw


# 1. Install via ClawdHub
clawdhub install trunkate-ai
 
# 2. Configure for Silent Optimization
export TRUNKATE_API_KEY="tk_live_12345..."
export TRUNKATE_AUTO_BUDGET="20%"  # Always-on Smart Buffer
 
# 3. Work normally — Trunkate manages context in the background.
# (Optional: call manually for a specific focus pivot)
openclaw "Optimize my history for the current task"

REST API


curl -X POST https://api.trunkate.ai/optimize \
  -H "Authorization: Bearer tk_live_1234567890abcdef" \
  -H "Content-Type: application/json" \
  -d '{
    "text": "Your very long context...",
    "task": "Summarize this document",
    "budget": "20%",
    "model": "gpt-4o"
  }'

Next Steps

Now that you’ve run your first optimization, learn how to refine the output further.

Integration Methods: Dive deep into the specific SDK or tool for your stack.
Authentication Guide: Learn how to manage your keys.
Intent Normalization: See how we remove fluff.
Web Playground: Test your optimized prompts.