Setup and Installation

Create an account and install the SDKs

Overview

Welcome to Confident AI! Getting set up takes a couple of minutes. You'll need to:

Create a free account and grab your project API key
Install the SDK for what you want to do — DeepEval for running evals locally, confident-trace for tracing your production app, and the Confident AI SDK for working with Confident AI from code
Point the SDK at your project with CONFIDENT_API_KEY (CONFIDENT_PROJ_API_KEY for the Confident AI SDK)

Rolling this out for your team?Plan the fastest path from setup to a production evaluation workflowTalk to us

Create an account
Navigate to cloud.confident-ai.com and create a free account. Once logged in, you'll be able to access your project API key from the dashboard.
Set API key as env variable
Set your project API key as an environment variable:
```
export CONFIDENT_API_KEY="confident_us..."
```
Both DeepEval and confident-trace read this same variable. For Confident AI SDK, set your project API key as CONFIDENT_PROJ_API_KEY:
```
export CONFIDENT_PROJ_API_KEY="confident_us..."
```
If you're using DeepEval, you can alternatively login through the CLI:
```
deepeval login
```
This will prompt you to enter your API key and save it locally.
Install the SDK
At the root of your project directory, install the package you need. Requires Python 3.10+.
pip install -U deepeval
pip install confident-trace
pip install confident-ai
Not sure which one? See the note below — most teams end up installing both.

Create an account
Navigate to cloud.confident-ai.com and create a free account. Once logged in, you'll be able to access your project API key from the dashboard.
Set API key as env variable
Set your project API key as an environment variable:
```
export CONFIDENT_API_KEY="confident_us..."
```
Or add it to your .env file:
```
CONFIDENT_API_KEY=your-api-key-here
CONFIDENT_PROJ_API_KEY=your-api-key-here
```
Both DeepEval and confident-trace read CONFIDENT_API_KEY, and the Confident AI SDK reads the same key from CONFIDENT_PROJ_API_KEY.
Install the SDK
At the root of your project directory, install the package you need. confident-trace requires Node.js 22+ (and tsx if you run TypeScript source directly).
npm install deepeval
npm install confident-trace npm install -D tsx
npm install confident-ai
Not sure which one? See the note below — most teams end up installing both.

Which package do I need? Confident AI has three SDKs, and they all use your project API key:

DeepEval is our open-source evals framework. Use it to run evals locally with metrics, and run unit tests in CI/CD. Start with the evaluation quickstart.
The Confident AI SDK (confident-ai) is the typed client for the Confident API. Use it to manage datasets and prompts, run evals on Confident AI against a metric collection, and gate deployments with confident-ai gate. It reads the key from CONFIDENT_PROJ_API_KEY. Start with the SDK reference.
confident-trace is our OpenTelemetry-native tracing SDK. Use it to instrument your production LLM app so traces, spans, and threads show up in the Observatory — call init() once at startup and it auto-instruments the providers and frameworks it finds. Start with the tracing quickstart.

You don't have to choose up front. Evals on production traces (online evals) run on the platform, so once your traces are flowing you can score them without any extra DeepEval code.

Confident AI's data lives in the US by default, but if you're a customer with a data residency in the EU, you'll need to point the SDKs at our EU servers and databases instead. DeepEval and the Confident AI SDK use the API base URL, while confident-trace uses the OTEL endpoint:

export CONFIDENT_BASE_URL="https://eu.api.confident-ai.com"
export CONFIDENT_OTEL_ENDPOINT="https://eu.otel.confident-ai.com/v1/traces"

This step is only required if you've created an account in the EU data region. Note that CONFIDENT_BASE_URL on its own does not change where confident-trace sends traces — you need both variables. See data residency for details.

If you're on a self-hosted deployment, your API key does not point the SDKs anywhere useful on its own — regions only switch between our managed hosts, so you must set the base URL (for DeepEval and the Confident AI SDK) and OTEL endpoint (for confident-trace) to your own deployment:

export CONFIDENT_BASE_URL="https://api.yourdomain.com"
export CONFIDENT_OTEL_ENDPOINT="https://otel.yourdomain.com/v1/traces"

See setting the base URL to your deployment for the POC ports and the OTEL endpoint.

Next Steps

Now that you have Confident AI set up, the next section will guide you through LLM evaluation in development. This is where you'll learn to systematically test and improve your AI applications before deployment.

LLM Evaluation

Run your first evaluation with DeepEval and see test results on Confident AI.

LLM Tracing

Instrument your app with confident-trace and see your first trace in the Observatory.

Last updated on October 23, 2026

Overview

Login with your API key

Create an account

Set API key as env variable

Install the SDK

Create an account

Set API key as env variable

Install the SDK

Next Steps