Getting Started

Get up and running with Risicare in minutes.

Risicare is in closed beta

Access is by invitation. The link in your invitation email opens the sign-in page — start there rather than from the addresses on this page. The link does not sign you in by itself: sign in with the same email address the invitation was sent to, or it is not accepted. If a step here does not work for you during the beta, write to support@risicare.ai.

This section takes you from an invitation to your first trace: sign in, create an API key, install the SDK, and send a trace.

Quick Links

Quickstart

Start tracing in 5 minutes

Learn more

Installation

Install and configure the SDK

Learn more

Core Concepts

Understand traces, spans, and agents

Learn more

What is Risicare?

Risicare is an observability platform for AI agents. It provides:

Automatic Instrumentation - Trace LLM calls with import risicare and two environment variables in Python, or with init() and a patchX() call in JavaScript
Decision-Level Observability - Track THINK/DECIDE/ACT phases
Error Classification - An exception that leaves a traced call gets a code from a 154-code taxonomy. The SDK assigns 19 of the codes, the same in Python and JavaScript (details)

What runs today

Tracing and error classification run today. Root-cause diagnosis and fix recommendations are built and held for the beta: a diagnosis cannot be requested and automatic diagnosis is off, so no new diagnosis or fix recommendation is produced. There is no way to read earlier diagnoses or fixes with an API key during the beta, and the dashboard does not show them. Canary and A/B deployment of fixes are built and gated off: a fix cannot be promoted and a deployment cannot be created. Hypothesis testing is built but not connected to the runtime.

Progressive Integration

Risicare uses a tiered approach that lets you start simple and add depth as needed:

Tier	What You Add	What You Get
0	`import risicare` + `RISICARE_API_KEY` + `RISICARE_TRACING=true`	LLM calls of the supported providers traced
1	`risicare.init()`	Explicit configuration
2	`@agent()`	Agent identity tracking
3	`@session`	User session grouping
4	`@trace_think/decide/act`	Decision phase tracking
5	`@trace_message/delegate`	Full multi-agent support

How It Works

Your Agent Code
      ↓
  SDK Instrumentation (automatic) — classifies errors as they are recorded
      ↓
  Rust Gateway (batching, validation)
      ↓
  ClickHouse (traces) + PostgreSQL (configs)
      ↓
  Diagnosis Engine (4-stage LLM pipeline) — held for the beta
      ↓
  Fix Generation — held (it runs only after a diagnosis)
      ↓
  Hypothesis Testing → Deployment (built — gated off)

Next Steps

Get invited and sign in

Access is by invitation — there is no self-serve sign-up. Sign in at app.risicare.ai with GitHub, using the account whose email matches your invitation — the public email on your GitHub profile if you have set one, otherwise your primary email. The Quickstart shows where your API key is shown.

Install the SDK

Run pip install risicare and set RISICARE_API_KEY to your key. A key is shown once, when it is created — copy it then.

Start Tracing

Add import risicare to your entrypoint, set RISICARE_TRACING=true, and run your agent.

View Dashboard

Open app.risicare.ai to see traces.

Review Classified Errors

Failed calls are flagged and classified against the error taxonomy. Root-cause diagnosis and fix recommendations are held for the beta.

Edit this page on GitHub

NextQuickstart