Introduction
Helicone is the open-source LLM observability platform for developers to monitor, debug, and improve production-ready applications.
Log your LLM requests, evaluate, and experiment with prompts, and get instant insights that helps you push changes to production with confidence. Helicone is the CI workflow designed for the entire LLM lifecycle.
Quick Start
Welcome to Helicone! Integrate your preferred provider with Helicone in seconds.
Create an account
Once you have an account, proceed to the next step.
Generate an API key
For more details on Helicone API keys, check out the Helicone Auth docs.
Pick your preferred integration method
Select the provider you are using below as the next instruction varies.
Send your first request 🎉
Once we receive your requests, you will see them in the Requests tab.
You will also see that your Dashboard has been updated with your new request.
What’s next?
We curated a list to help you make the most of Helicone, but you’re welcome to explore the product on your own!
Add a custom property
Label your requests to segment, analyze, and visualize them.
Create your first prompt
Version your prompt and inputs as they evolve.
Track a session
Group and visualize multi-step LLM interactions (i.e. AI agents).
Cache responses
Cache and watch how much time and cost you saved.
Recommendations
The following guides are optional, but we think you’ll find them useful.
Run a prompt experiment (new)
Did tweaking your prompt or using a different model cause any regression? The answer is in Experiments.
Bring it to PostHog
Helicone has teamed up with PostHog to bring your LLM analytics closer to all your other dashboards.
Explore Features
Discover all the features for monitoring and experimenting with your prompts.
Prompts
Manage and version prompts from code or UI.
Sessions
Trace agentic workflows and visualize them.
Experiments (new)
Tune your LLM prompts for production.
Webhooks
Instantly react to events, trigger actions, and integrate with external tools.
Custom Properties
Label and segment your requests.
Caching
Save cost and improve latency.
Scores
Evaluate prompt performance and quantify model outputs.
Omit Logs
Remove request and responses.
User Metrics
Get insights into your user’s usage.
Datasets & Fine-tuning
Curate datasets and fine-tune your LLMs.
Feedback
Provide user feedback on output.
Gateway Fallback (Beta)
Utilize any provider through a single endpoint.
Retries
Smartly retry requests.
Rate Limiting
Easily rate-limit power users.
Key Vault
Manage and distribute your provider API keys securely.
Moderation Integration
Integrate OpenAI moderation to safeguard your chat completions.
LLM Security
Secure OpenAI chat completions against prompt injections.
Webhooks
Set up integrations tp subscribe Helicone events.
Customer Portal
Easily manage your customers and their usage.
Further Reading
Proxy or Async?
Determine when you should use a proxy or async function in Helicone.
How we calculate costs
A detailed breakdown of our process to calculate cost per request.
Understanding Helicone Headers
Every header you need to know to access Helicone features.
Questions?
Although we designed the docs to be as self-serving as possible, you are welcome to join our Discord or contact help@helicone.ai with any questions or feedback you have.
Interested in deploying Helicone on-prem? Schedule a call with us.