Auto EEG Analysis (AEA)

AEA provides reusable EEG analysis skills and study recipes for large language model coding agents. MNE-Python is the primary computation backend. Selected EEGLAB and FieldTrip routes are described in the bundled documentation.

This package installs the resource library and a project initialization command. It includes 22 skills, 10 study recipes, analysis helpers, templates and setup documentation. It does not install or configure a model client or an account.

Install and create a project

Use Python 3.10 or newer in an environment of your choice.

python -m pip install auto-eeg-analysis==0.3.3rc1
aea-eeg init ./my-eeg-study --agent codex
cd my-eeg-study

Use --agent claude for Claude Code or --agent both for both clients. The command copies the library into a new or empty directory and creates project-local skill links using AEA's existing installer. Existing nonempty directories are refused. It does not change global client configuration. python -m auto_eeg_analysis accepts the same arguments as aea-eeg.

The default installation has no scientific runtime dependencies. Reuse your existing MNE environment, use the included environment.yml to create one, or install the analysis dependencies from the existing environment specification.

python -m pip install 'auto-eeg-analysis[analysis]==0.3.3rc1'

Read AGENT_GUIDE.md, select procedures through RESOURCE_INDEX.md, and approve the study-specific analysis plan. Start the separately installed coding client from the initialized project directory. aea-eeg path shows the installed resource directory.

Version and evidence

This packaging candidate uses published AEA source commit cdc0c30105cc4547aed8e5ece41739a8462cb47a. The package version identifies this distribution. Retained scientific validation records and plugin versions keep their original identifiers. Packaging checks do not extend the numerical validation coverage in docs/CERTIFICATION_LEVELS.md.

The package contains no participant datasets, model credentials or archived model sessions. The Git repository remains the development source and the place for contributions and support.

License and citation

Code and skills use the repository's MIT license. Study recipes use CC BY 4.0 as stated in recipes/README.md. Preserve the recipe citations and original license notices. CITATION.cff retains the source repository's citation metadata.

Project · Setup guide · Validation coverage