The Self-Alignment Framework

Grounded in classical philosophy to establish deterministic boundaries for autonomous AI agents.

SAFi
The Math

The Framework at a Glance

In Thomistic philosophy, Synderesis is the innate habit by which the intellect apprehends the first principles of moral reasoning. In SAFi, we adapt this concept as the foundational compiler that turns an organization's charter, policies, values, and rules into the structured standards that govern the agent's execution faculties.

Throughout SAFi, Synderesis and values are sometimes used interchangeably, because values are what Synderesis produces. Synderesis is the mechanism. Values are the compiled standards the mechanism enforces.

Click any card for more detail.

1 Synderesis The compiler of the framework. Turns the charter, policies, values and rules into the standards the whole loop operates under. Synderesis 2 Intellect The guide to discernment and understanding. Drafts the answer and proposes tool calls, without the rubrics it will be graded against and with no power to act. Read Intellect → 3 Will The power to act. The disciplined executive that turns intention into action. Read Will → 4 Conscience The inner guide. Independently judges whether an action stayed aligned with the values. Read Conscience → 5 Spirit The manifestation of alignment. Integrates the whole over time and surfaces drift. Read Spirit →

What the Framework Is

SAF is a framework grounded in classical philosophy. It describes how values, understanding, choice, judgment, and coherence work together to produce aligned agents.

Download Original Manuscript (PDF)

SAFi is the software implementation of the framework. It is an open-source runtime governance engine for AI agents. Read about SAFi →

The SAFi Loop

Every agent in SAFi follows this loop. It is what makes the process auditable and repeatable. Nothing here is specific to chat: a turn can be started by a person, by an API call or by a schedule, and it can act on the world and not only answer.

  1. Trigger: a user prompt, an API call, or a scheduled task
  2. Phase 0: screens the incoming prompt for injection, PII and scope probes, before any model is called
  3. Intellect: drafts an answer, or proposes a tool call, and can execute nothing
  4. Will, tool intent: authorises a proposed tool call by its arguments and not just its name, and the call runs only if it passes
  5. Will, structure: checks the draft, never the tool call, for its required structure
  6. Conscience: scores the draft value by value, with a reason and a confidence
  7. Will, hard gate: enforces the gates, and approves, blocks or redirects
  8. Spirit: folds the scores into longer-term alignment and measures drift
  9. Will, spirit: rules on the resulting alignment score
  10. Answer: returned, with the governance record written for the turn

Try the live demo. Log in as Guest and you will land on the Fiduciary agent by default. It has access to real-time stock data, and it will refuse to give you financial advice.

You can also try it without leaving this page. The chat button in the corner is a governed SAFi agent answering from this framework’s own documentation.

Continue Reading

How SAF Was Born: The Philosophical Foundation of SAFi

Where the framework came from, and why it was built on classical philosophy rather than invented from scratch.

What is an AI Agent? Beyond the Chatbot

It is kind of surprising that there is still no standard definition for an AI agent in the industry. You will often find people debating what an agent actually is. If you step back and think about whe...

AI in Plato's Cave

Expectations about AI are high right now. Many people, especially software engineers and developers, are scared that their jobs will be automated away. But I've been in IT for over twenty years, so th...

I Don't See the Value in Frontier AI Models Anymore

After building SAFi with increasingly capable coding tools, I found that a smaller, cheaper model was enough for most of my practical engineering work. The right model, architecture, and boundaries matter more than raw capability.

All articles

The architecture, the concepts, the applications, and the arguments behind them.