Your AI. Your infrastructure. Your governance.
AI agents are becoming capable of doing real work. They can research, analyze information, interact with systems, make decisions, and execute business processes. But giving an AI agent access to real systems creates a different problem.
"How do you let an AI do useful work without giving up control?"
That's the problem RunSAFi exists to solve. We help organizations build and run AI agents they can trust with real work. We bring the tech. You bring the resources and use cases.
Our mission
Our mission is to empower organizations to harness the full potential of AI without ever compromising control over their data, infrastructure, or code.
Core values
Transparency
Everything we build is either available under an open-source license or deployed directly in the client's environment without restrictions. We don't believe clients should have to trust a black box. Our technology is built for review, inspection, and understanding.
Institutional Alignment
AI should not operate based on a model provider’s generic safety guidelines. It should align with your organization’s specific mission, core values, and policies. We build systems that enforce your defined values at runtime, ensuring the AI acts as a true agent of your institution, augmenting your team’s judgment, never overriding your governance.
Data Sovereignty
Organizations should remain in control of their AI. They should be able to decide what their AI can access, what it can do, where it runs, what information it uses, and what happens to their data. We build our solutions so that organizations retain that control rather than surrendering it to an AI provider.
Built from an infrastructure mindset
RunSAFi was built from more than two decades of enterprise IT experience. Traditional infrastructure operates on principles that are easy to understand:
- Least privilege
- Defined boundaries
- Separation of responsibilities
- Auditability
- Controlled access
- Human accountability
AI agents introduce a different kind of system. They can interpret information, reason about it, use tools, and take actions. Their capabilities are powerful, but their behavior is not something you should govern with prompts alone.
We believe AI agents need an operational governance layer that sits between what an agent wants to do and what it is actually allowed to do.
That's where SAFi comes in.
SAFi: Governance at runtime
SAFi is the open source governance engine behind RunSAFi. It provides the governance layer that evaluates an agent's reasoning and actions against defined policies before allowing those actions to proceed.
The goal isn't to make AI predictable. The goal is to make the boundaries around AI predictable.
- Enforce defined policies
- Control tool access
- Require human approval
- Maintain an auditable record of what the agent did and why
That allows organizations to use capable AI systems while maintaining control over how those systems operate.
We believe AI should operate where your business operates
Your data is yours. Your infrastructure is yours. Your business processes are yours. Your AI should not require you to surrender control of them.
That's why we design RunSAFi deployments to operate within your environment whenever the use case requires it. You control the infrastructure, the data, and the model credentials.
We bring the technology and the expertise to build, deploy, secure, monitor, and maintain the system.
And when your team is ready, the system can be handed over to your organization.
From framework to working systems
RunSAFi is part of a larger project. The distinction matters as much as the connection.
The Self-Alignment Framework
The philosophy. The Self-Alignment Framework explores how values, reasoning, intention, judgment, and action can be structured into a framework for governing autonomous systems.
SAFi
The technology. SAFi turns those ideas into an open source governance engine for AI agents.
RunSAFi
The service. RunSAFi brings SAFi into real organizations. We build, deploy, and operate AI agents around real business use cases.
SAF is the framework. SAFi is the engine. RunSAFi is the business.
Meet the founder
RunSAFi was founded by Nelson Amaya, the original author of the Self-Alignment Framework and the architect of its open-source implementation, SAFi (Self-Alignment Framework Interface).
After a 22-year career in enterprise IT spanning roles from the help desk to IT Director, managing complex cloud environments, and leading service desk operations, Nelson built RunSAFi to solve a specific problem: how to deploy highly capable AI agents that align with an organization's mission and values while strictly adhering to enterprise security and regulatory compliance standards.
We're not here to replace your IT team
We're here to help your organization put AI to work. You know your business. You know your systems. You know where the valuable work is.
We bring the technology and engineering expertise to turn those opportunities into working AI agents and deploy them within the boundaries your organization requires.
You bring the resources and use cases. We bring the tech.
Built for real work
We don't believe the future of AI is simply better chatbots. The interesting question is what happens when AI starts doing actual work.
- Researching
- Analyzing
- Processing
- Updating systems
- Coordinating workflows
- Making recommendations
- Taking actions
That's where AI becomes genuinely useful. And that's also where governance becomes essential. RunSAFi exists to help organizations get there without giving up control.
Let's build something useful
Have a business process you think an AI agent could handle? We'd like to hear about it.
Your AI. Your infrastructure. Your governance.