empirica
Make AI agents and AI workflows measurably reliable. Epistemic measurement, Noetic RAG, Sentinel gating, and grounded calibration for Claude Code and beyond
What is empirica?
Make AI agents and AI workflows measurably reliable. Epistemic measurement, Noetic RAG, Sentinel gating, and grounded calibration for Claude Code and beyond
Overview of empirica
What it is
Empirica is an open-source epistemic infrastructure framework for AI agents and workflows. It measures the gap between what an AI predicts and what is true, providing memory, measurement, and calibration across sessions. The framework sits beside AI models to track what they know, gate what they do, and compound learning across session boundaries. It has no cryptocurrency, token, coin, or blockchain component.
Key features
- Measures before acting: AI investigates the codebase before making edits, with a Sentinel gate that blocks changes until understanding is demonstrated.
- Four-layer memory with provenance and confidence: findings, dead-ends, assumptions, and mistakes persist, are retracted or superseded when wrong, and are retrieved by relevance to the task.
- Domain-aware CHECK gate: uses thresholds scaled by criticality and tightened by the calibration of the model doing the work.
- Real-time confidence display: a live statusline in the terminal shows stage, confidence, goals, unknowns, assumptions, and learning marks.
- Three-vector calibration model: self-assessed, observed (from deterministic checks), and AI-reasoned grounded state with rationale.
- Temporal entity model: auto-extracts functions, classes, and imports from every file edit to track what is alive and stale.
- Optional cross-AI mesh via Cortex: peer AIs propose work, accept/decline, and completion handshakes carry commit SHAs.
Use cases
Empirica is designed for developers and teams using AI agents (such as Claude Code) to write and ship code, run pipelines, research, and coordinate with other agents. It is used to make AI workflows measurably reliable by providing an audit trail of what the AI believed, assumed, or checked, and by gating autonomous actions based on demonstrated understanding and calibration.
Getting started
bashpip install empiricaempirica setupFor alternative installation methods (Homebrew on macOS, Docker) and detailed setup instructions, see the repository README.
License and pricing
Empirica is licensed under the MIT License and is open source.
Based on the project's public README and repository metadata.