Skip to main content
Follow these steps to get your first results. Use the dashboard, Ask Oqo, MCP, or the CLI.
1

Get access

Go to oqoqo.ai and select Get started. Sign in, then open a project. Organizations start on the free plan, which resets runs each month. See Pricing.
2

Connect providers

Open Organization settings → Model providers. Connect an API key, OpenRouter, Vercel AI Gateway, or another compatible provider. See Model providers.
3

Assign Oqoqo AI features

Open Organization settings → Oqoqo AI features. Assign models for Ask Oqo, Rubric generation, Run evals, Run insights, and Experiment insights. See Oqoqo AI features.
4

Enable agents

Enable agents under Manage → Agents. For Claude Code, use an Anthropic API key or Anthropic (Claude Subscription). For Codex, use an OpenAI API key or OpenAI (ChatGPT Subscription). For Cursor, connect an API key from a paid Cursor plan. For Grok Build, connect an xAI API key. For Qwen Code, connect Alibaba Model Studio or another compatible organization provider. For Antigravity, connect a Google Gemini API key. See Supported agents.
5

Connect MCP or CLI

Install the recommended plugin for your coding agent, or install the CLI. Your agent can then create tasks, launch experiments, and read results.
6

Create a task

Write clear instructions and a few pass or fail requirements. Attach any Files the agent needs. Keep Machine on the Standard Dev Machine unless the work needs more.
Files and the machine belong to the task. See Machines and files.
7

Create treatments

Keep a baseline treatment (raw agent). Create a second treatment with your skill, MCP server, CLI, or SDK attached. See Treatments.
8

Launch

Select New experiment, or start from a Home template such as Getting started. Choose tasks, agents, and treatments. Set trials. Review Max Concurrency. Turn on Evals and Insights. Select Launch. See Launch options.
9

Read results

Use Matrix for pass rate and lift. Use Frictions for where agents got stuck. Open any run for Traces, Output, and Evals. Ask Ask Oqo for a plain-language summary.
Experiment Matrix with search, metric controls, and result cells.
For terms, see Core concepts. For setup problems, see Troubleshooting.