Quickstart

1. Connect a model provider

In the app, go to Model Gateway → Providers and connect a provider (OpenRouter, OpenAI, or an OpenAI-compatible endpoint). You bring your own key - model spend bills to it. Agents reference models as provider/model, where the first segment picks the connection and the rest is passed through. An OpenRouter model keeps its own vendor prefix and gains one: openrouter/openai/gpt-4o-mini. See Models & providers.

2. Create an API key

Go to API keys and create one. Keep it safe; it is shown only once.

export CRUQAI_API_KEY=cruq_sk_...
export CRUQAI_BASE_URL=https://app.cruq.ai

3. Install the CLI

pip install cruqai
cruqai login   # or rely on the env vars above

4. Run an agent

Every new workspace ships an Example assistant. List your agents, then run one:

cruqai runs                 # recent runs
cruqai run <agent-id> "What is 15% of 240?"

You will see a live tool-call trace as the agent works, and the run is persisted for later inspection:

cruqai trace <run-id>

5. Ground it in your own knowledge

cruqai knowledge create "Product docs"
cruqai knowledge add-doc "Product docs" --title "Policies" --file policies.pdf
cruqai knowledge search "refund policy"

Wire a knowledge_search tool into an agent and it will retrieve grounded context at run time. See Knowledge.

6. Find out which model to use

Rather than guessing, run the same test cases across several models and compare them on score, latency, and cost:

cruqai evals create-suite "regression" --agent <agent-id>
cruqai evals add-case regression --name "does the math" \
  --input "What is 15% of 240?" \
  --scorers '[{"type":"contains","value":"36"}]'
cruqai evals run <agent-id> --suite regression \
  --model fast=openrouter/openai/gpt-4o-mini \
  --model smart=openrouter/anthropic/claude-3.5-sonnet --watch

See Evals.

Getting Started

Concepts

Reference

1. Connect a model provider

2. Create an API key

3. Install the CLI

4. Run an agent

5. Ground it in your own knowledge

6. Find out which model to use

​1. Connect a model provider

​2. Create an API key

​3. Install the CLI

​4. Run an agent

​5. Ground it in your own knowledge

​6. Find out which model to use

1. Connect a model provider

2. Create an API key

3. Install the CLI

4. Run an agent

5. Ground it in your own knowledge

6. Find out which model to use