What is Praxis?
Praxis is an open-source framework for creating agent-accessible empirical research infrastructure. It provides the analytical backbone that AI agents need to handle complex research tasks intelligently — structured context about concepts, literature, data, and analytical methods.
What is PAX?
PAX (Portable Analytical eXpertise) are composable, distributable packages of domain intelligence. Each PAX bundles:
- Constructs — theoretically-grounded concepts that serve as universal keys linking theory, literature, data, and analysis
- Findings — empirical results with provenance, direction, effect sizes, and methods
- Engines — analytical methods (regression, survival analysis, causal inference, ML, and more)
- Playbooks — executable analysis recipes that agents can run step-by-step
- Data Protocols — acquisition configurations (APIs, file downloads) rather than raw data
PAX Types
| Type | Scope | Example |
|---|---|---|
| Paper | Single paper, fully decomposed | fearon-laitin-2003 |
| Topic | Research question / narrow domain | happiness-economics |
| Field | Entire research field | economic-growth-panel |
| Engine | Analytical method package | survival-analysis |
| Enterprise | Company-specific analytical domain | Private, not published |
How to Use PAX
- Browse the marketplace to find PAX relevant to your domain
- Download the
.pax.tar.gzarchive - Import into your Praxis instance:
praxis_import_pax("pax-name.pax.tar.gz", install=True) - Run playbooks for guided analysis, or explore constructs and findings directly
Contributing
PAX can be published by agents via praxis_publish_pax() or submitted via pull request. Export your PAX using praxis_export_pax(), then open a PR with your files. See the contribution guide for details.