AIBuildAI Science Guide

1. Register and subscribe

Create an account at accounts.aibuildai.io/sign-up. After signing in, go to the Billing section and switch to the Science plan ($1000 / month). Once payment is processed, your account will show an active Science subscription.

2. Install

AIBuildAI requires a Linux x86_64 machine (Ubuntu 20.04 or newer).

curl -fsSL https://raw.githubusercontent.com/aibuildai/AI-Build-AI/main/install.sh | AIBUILDAI_LINE=science sh

3. Log in

Sign in to your AIBuildAI plan:

aibuildai login

To verify your login status, run aibuildai whoami — the output should show an active Science plan and confirm that runs are authorized.

4. Sign in to Claude Code or set your API key

If Claude Code is already signed in on this machine, no Anthropic API key is needed. AIBuildAI automatically uses the local Claude Code credentials.

If Claude Code is not signed in, run:

claude auth login

You can also use an API key instead. On this edition the key always goes in AIBUILDAI_API_KEY, whichever model provider you configure:

export AIBUILDAI_API_KEY=your-api-key

The model you select in task.yaml decides which provider's key is expected:

• Anthropic — the default. Use your Anthropic API key.
• OpenAI — set a gpt-* model in the config. If Codex is signed in, that login is used; otherwise put your OpenAI key in AIBUILDAI_API_KEY.
• DeepSeek — set a deepseek-* model in the config and use your DeepSeek key.

Add the export line to your shell profile to keep it set across sessions.

5. Run

AIBuildAI Science is driven by a YAML config, not by command-line flags. You write a config file and pass that file to aibuildai run:

aibuildai config > task.yaml    # writes a starter config with every field
# edit task.yaml: set run.task_name, run.data_root, run.playground_root
aibuildai run task.yaml

Where your data goes

On this edition run.data_root is the task folder itself: the directory holding what the task means plus every material the task needs. The run reads that folder whole and expects no layout inside it. If the path does not exist, the run stops before any model call with no task folder found at ....

Try the worked example

The repository ships tasks/protein-ec-prediction-science.yaml, a config for predicting the enzyme class (EC number) of a protein from its amino acid sequence, together with its dataset. Clone the repo, point run.playground_root at a directory of your choice, and run it directly:

git clone https://github.com/aibuildai/AI-Build-AI.git
cd AI-Build-AI
# Edit tasks/protein-ec-prediction-science.yaml first: set run.playground_root
# to a writable directory of your choice. It ships as a placeholder.
aibuildai run tasks/protein-ec-prediction-science.yaml

Other commands: aibuildai resume to resume a stopped run, aibuildai memorize to summarize past runs into memory, aibuildai replay <run-dir> to replay a finished run, and aibuildai --help for the full list. For every configuration option, see github.com/aibuildai/AI-Build-AI.