If you’re interested in getting set up, check out the dashboard guide.

​
Welcome to Your Agent’s Control Room

AI agents don’t just make predictions — they perform multi-step workflows, invoke tools, navigate interfaces, and adapt in real-time. That complexity comes at a cost: it’s easy for agents to behave unexpectedly, fail silently, or succeed for the wrong reasons. We built this platform to give you deep observability into how your agent thinks, acts, and learns — across every run.

​
Why Observability Matters

Running an agent once doesn’t tell you how it behaves.
  • Agents are non-deterministic
  • They’re made of chained actions, not just outputs
  • A single successful session can hide dozens of edge cases
To build agents you can trust, you need to see what they’re doing — and why.

​
What This Platform Does

We help you:

​
Automatic Instrumentation

Lucidic is designed to be as non-intrusive as possible. With just two lines of code, you get:
  • Auto Event Creation: LLM calls from supported providers (OpenAI, Anthropic, LangChain, PydanticAI) are automatically captured
  • Auto Step Creation: If no step exists when an LLM call occurs, one is created automatically
  • Auto-Ending Steps: Steps are automatically ended when your session ends
  • Auto Session Ending: Sessions automatically end when your script exits via an atexit handler
  • Auto Cost Tracking: Token usage and costs are calculated automatically for each LLM call
  • Auto Error Handling: Failed operations are tracked without breaking your workflow

​
How It Works

When you initialize Lucidic with a provider: 
import lucidicai as lai
lai.init(providers=["openai"])
The SDK:
  1. Instruments the provider using OpenTelemetry under the hood
  2. Intercepts all API calls to track inputs, outputs, tokens, and timing
  3. Creates steps and events automatically as your code runs
  4. Handles cleanup gracefully even if your script crashes
This means you can add Lucidic to existing codebases without any refactoring - just add the two lines at the top of your script and you’re done! 
import lucidicai as lai
lai.init(providers=["openai"])  # That's it! All OpenAI calls are now tracked

​
Next Steps

Ready to dive in? We’re excited to help you build agents that are observable, debuggable, and production-ready.

​
Need Help?

Questions, feedback, or something not working right? Reach out to us at team@lucidic.ai — we’d love to help.