ARF is a production‑ready reliability and governance layer for AI systems, combining Bayesian risk scoring, semantic memory, auditability, and self‑healing loops.
The framework is founder‑controlled and pilot‑first – the core engine is protected, access‑gated, and offered to selected enterprises under outcome‑based pricing.
⚠️ Important – The core ARF engine (agentic_reliability_framework,arf-api,enterprise) is not open source. It is proprietary, access‑controlled, and available only to qualified pilots and enterprise customers.
- Observes agent behaviour and system health
- Detects anomalies and failure patterns using Bayesian inference
- Applies governance and policy controls (approve/deny/escalate)
- Supports adaptive healing and recovery (self‑healing loops)
- Stores incidents, outcomes, and learnings for future decisions (semantic memory)
| Repository | Status | Description |
|---|---|---|
arf-spec |
Public | Canonical specification – data models, API contracts, decision rules (Apache 2.0) |
arf-frontend |
Public | Demo dashboard UI (sanitised, uses mock data) |
pitch-deck |
Public | Public overview and vision |
agentic_reliability_framework |
Private | Protected core engine – Bayesian risk scoring, expected loss minimisation |
arf-api |
Private | Production control plane – FastAPI service |
enterprise |
Private | Customer‑specific adaptations, proprietary heuristics |
research |
Private | Experimental methods and internal models |
The core ARF engine is not publicly available. Access is granted only to:
- Selected pilot customers – time‑limited free trial under a mutual qualification agreement
- Enterprise licensees – outcome‑based pricing tied to measurable risk reduction
To request pilot access, email [email protected] with:
- Use case description
- Expected monthly incident volume
- Cloud environment (AWS, Azure, GCP, on‑prem)
ARF pricing is not per‑seat or per‑request. It is directly tied to the amount of operational risk the system removes from your AI workflows.
We measure risk reduction via auditable pre/post Bayesian scores – you pay only for verified improvement.
We accept limited contributions to public repositories (arf-spec, arf-frontend, pitch-deck) – bug fixes, documentation, demo improvements.
We do not accept pull requests against private core repositories.
- Open an issue describing your proposed change.
- Wait for a maintainer to assign the issue.
- Sign a Contributor License Agreement (CLA) if requested.
- Submit a pull request referencing the issue.
All changes are reviewed and merged at the founder’s discretion.
- Public repositories (
arf-spec,arf-frontend,pitch-deck) are licensed under Apache 2.0. - The core engine and all private repositories are proprietary – not licensed for public use.
- 📬 Email: [email protected]
- 💬 Slack: Join workspace
- 🔗 LinkedIn: Juan Petter
- 📅 Book a call: Calendly
Stewarded by the founder – pilot‑first, outcome‑based pricing.