Find what's wrong with your AI — and fix it.

The AI Decision Engine for improving production systems. Automatically surface plain-English insights, detect critical paths, and optimize costs for your AI agents and chains.

$pip install trackly
DevDevDevDevDev
+1.2k

Join 1,200+ developers building the future of AI

Questions? support@tracklyai.in

0
Events
0
Tokens
6
Providers
2
Lines of code

Install in 2 lines of code

main.py
from trackly import Trackly from langchain_anthropic import ChatAnthropic # 1. Init with metadata trackly = Trackly( feature="chatbot", environment="prod" ) # 2. Attach callback — that's it llm = ChatAnthropic( model="claude-3-5-sonnet-latest", callbacks=[trackly.callback()], ) # Every Claude run is now analyzed automatically ✓

Detect every
critical path

Surface what's wrong with your AI instantly. Trackly automatically detects critical paths, highlighting the slowest, most expensive, and failure steps in every agent run — so you can fix them.

Live Critical Path

Automatically highlight the slowest and most expensive steps in every run.

Step-By-Step Analysis

Reconstruct every agent decision and model call into a logical, actionable map.

Auto Insights Engine

Surface plain-English findings on why a specific run was sub-optimal.

Features

Everything you need to fix
your AI systems

From zero-config provider detection to per-user cost attribution — Trackly gives you complete visibility without changing how you write code.

Auto Insights Engine

Stop reading dashboards. Trackly automatically surfaces plain-English findings from every run: "Cost spike caused by GPT-4 overuse in step 3" or "Latency bottleneck detected in the tool-calling loop."

OpenAIAnthropicGroqGeminiMistralOllamaTogetherFireworksBedrockCohere

Detected automatically

provider →openai
model →gpt-4o

Critical Path Detection

Automatically highlight the slowest, most expensive, and failure points. Visually tagged as the 🔥 Critical Path so you know exactly where to look.

summarizer
$18.42
chat
$13.71
rag-pipeline
$8.53
onboarding
$4.26

Cost Intelligence

Move beyond tracking. Get model efficiency suggestions: "Switch to GPT-4o-mini to save 68% cost."

312ms

▲ P95 · gpt-4o · last 24h

P50

198ms

P95

312ms

Run Comparison

Side-by-side diffs of cost, latency, and steps taken.

u_8821$4.21

u_3341$2.87

u_9102$1.44

u_0551$0.93

Live model pricing

Cost computed at ingest. Historical events stay accurate when providers change rates.

gpt-4oopenai$2.50 / 1M
claude-3-5-sonnetanthropic$3.00 / 1M
llama-3.3-70bgroq$0.59 / 1M
gemini-1.5-flashgoogle$0.075 / 1M

Smart Alerts

Every alert carries an interpreted reason for the spike.

Budget exceeded
summarizer hit $20 limit
Spike detected
3× usage in last hour

Daily usage trends

30-day rolling view. Spot spikes, monitor growth, export to CSV.

4.8M

+23% vs last month

Mar 1Mar 11Mar 22
How It Works

Activate the decision engine in 6 simple steps

From sign-in to your first automated AI insight — here's everything you need to get up and running with the Trackly Decision Engine.

1

Sign in & view your Dashboard

After signing in, you land on the Trackly Dashboard. If you don't have any projects yet, you'll be prompted to create your first one.

Step 1: Sign in & view your Dashboard
2

Navigate to Projects

Head to Organization Settings → Projects to see all your tracking environments. Each project isolates its data and API keys.

Step 2: Navigate to Projects
3

Create a new Project

Click "+ New Project", give it a name (e.g. "My App"), and hit Launch Project. Projects keep your tracking data cleanly separated.

Step 3: Create a new Project
4

Go to API Keys

Navigate to API Keys from the sidebar. You'll need a key to authenticate your SDK calls from your codebase.

Step 4: Go to API Keys
5

Generate an API Key

Click "+ Create API Key", give it a descriptive name, and optionally scope it to a specific project for tighter access control.

Step 5: Generate an API Key
6

Copy your API Key & start tracking

Your key is generated! Copy it immediately — you won't be able to see it again. Drop it into your environment variables and start tracking.

Step 6: Copy your API Key & start tracking
Pricing

Simple, transparent pricing

Start free. Scale when you're ready. No hidden fees — just raw performance and deep insights.

Starter

$0forever

For side projects and experiments.

1Mtokens / month
  • 3 Projects
  • 7-day data retention
  • Basic insights
  • Community support

Scale

$99/month

Advanced controls for high-volume traffic.

10Mtokens / month
  • Everything in Pro
  • 90-day data retention
  • Custom feature tags
  • Priority support
  • SLA guarantees

Need more? Contact us for enterprise pricing

About
Udaykumar Dhokia

Udaykumar Dhokia

Builder of Trackly · Software Developer

Trackly was built out of a real frustration. While working on agentic AI projects, I had no simple way to know how much each feature was costing me — especially when using Groq or Ollama where model names give no hint about the host platform.

I'm a Computer Engineering student from Ahmedabad with a focus on Full Stack Development, Agentic AI, and applied AI systems — RAG, vector search, and LLM tooling.

Agentic AILangChainFull StackRAGVector SearchFastAPIFlutterPostgreSQL
Contact

Got questions?
Let's talk.

Whether it's a bug, a feature request, a partnership idea, or just curiosity about how Trackly works — drop a message and I'll get back to you.

Usually responds within 24 hours