Skip to main content
This guide helps you choose the right tools and integration path based on how you use language models and agents.

​
Logging, gateway, or tracing?

​
LLM logging

Call the LLM yourself, and we help you monitor those calls.
If you already have LLM API calls in your code and just want observability (prompt, response, latency, user tracking, etc.), use the Logging API.
  • No latency
  • Works with any LLM provider
  • Fast setup with a single HTTP call
👉 Quickstart

​
LLM gateway

We help you to handle LLM calls and everything.
The Gateway is the easiest way to connect to major LLM providers. We can help you to handle rerouting, retries, load-balancing, caching, fallbacks, and etc..
  • 50 - 150ms latency
  • Support 250+ models
  • Built-in cost tracking, retries, observability
👉 Quickstart

​
Agent tracing

Track your agent workflows step by step
Perfect for multi-step chains, working with popular AI agent frameworks.
  • Native support for OpenAI Agent SDK, Mastra, and more
  • Custom step tracing with our SDK
  • Visualize execution trees, tools, retries, state changes
👉 Quickstart

​
Prompt management and evaluations

Isolate prompts from code, collaborate with prompt engineers, and iterate faster.
  • Version control for prompts
  • Human + LLM evaluation pipelines
  • Visual scoreboards, test sets, experiments
👉 Quickstart

​
Need help?

Join our discord — we’ll help you as best we can.
⌘I