Skip to main content

​
Trainly

The observability platform for agentic workflows. Trainly gives you full visibility into how your AI systems behave in production. Add a single decorator to your code and get tracing, scoring, anomaly detection, and version control — automatically.

​
Why Trainly

Trace

Capture inputs, outputs, latency, tokens, and cost with a single @observe decorator. No config needed.

Detect

Surface unusual patterns and behavioral changes across your AI interactions automatically.

Understand

Get plain-language explanations of what changed and why — no log-diving required.

​
Get started in 5 lines

from trainly import TrainlyClient
from openai import OpenAI

client = TrainlyClient(api_key="tk_...", project_id="proj_...")
openai_client = OpenAI()

@client.observe(model="gpt-4o", tags=["production"])
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

# Every call is now traced automatically
result = ask("What is AI observability?")

​
Explore the docs

Quickstart

Trace your first AI call in under 5 minutes.

Python SDK

Full guide to the @observe decorator, sessions, spans, and scoring.

React SDK

useTrainlyObserve hook, session providers, and pre-built components.

API Reference

REST API for trace ingestion, analytics, testing, and version management.