AI Agents Manifesto
A Framework for the Agent-First Era
Drafted by technology founders and practitioners in 2025
We are building the infrastructure for a fundamental shift in how work gets done. AI agents will handle an increasing share of cognitive tasks across every industry, just as automation transformed manufacturing. This isn't speculation—it's already happening in our codebases, customer service systems, and data pipelines.
Core Values
We have come to value:
Five Technical Realities
The fundamental truths shaping our technological landscape
Automation Will Scale Beyond Current Expectations
AI agents will handle complex cognitive work—financial analysis, code review, customer research, content creation, and decision-making. This isn't limited to simple task automation. We're building systems that can reason, plan, and execute multi-step workflows across domains.
Agent Engineering is Core Infrastructure
Understanding how to design, implement, and manage AI agents is becoming as fundamental as database design or API architecture. Teams that can't build and integrate agents effectively will struggle to compete with those that can.
This is a Platform Shift, Not an Enhancement
Just as mobile didn't just improve desktop computing but created entirely new business models, agents aren't just making existing work faster—they're changing what work looks like. Roles focused on routine analysis, research, and coordination will largely disappear.
Agents Require Unprecedented Data Access
Effective agents need broad context to make good decisions. This means access to customer data, internal communications, financial records, and operational metrics. This creates significant security and privacy challenges that traditional access control models weren't designed to handle.
Guardrails Are a Technical Requirement, Not a Nice-to-Have
Agents that can modify code, interact with customers, or make financial decisions need robust constraint systems. Without proper boundaries, agents can cause data breaches, make inappropriate decisions, or scale problems faster than humans can detect and correct them.
Agent Deployment Principles
Technical standards for organizations building agent systems
1. Scope-Limited Agent Design
- Each agent operates within clearly defined functional boundaries
- No general-purpose agents in production without extensive testing
- Agent capabilities should be composable but contained
- Regular capability audits and scope reviews
2. Data Access by Design
- Implement least-privilege access for all agent operations
- Maintain detailed audit logs of data access and modification
- Use encryption and access tokens rather than direct database access
- Regular access reviews and automatic permission expiration
3. Human-in-the-Loop Architecture
- Critical decisions require human approval or review
- Implement circuit breakers for agent systems
- Maintain rollback capabilities for agent actions
- Clear escalation paths for edge cases
4. Observable and Debuggable Systems
- All agent decisions must be traceable and explainable
- Implement comprehensive logging and monitoring
- Version control for agent behavior and training data
- A/B testing frameworks for agent performance
5. Continuous Security Assessment
- Regular penetration testing of agent systems
- Monitor for capability creep and unauthorized access
- Incident response plans specific to agent failures
- Security reviews for agent updates and modifications
The Future Demands New Skills
Understanding how to build, manage, and collaborate with AI agents isn't optional—it's the new literacy for the digital age. The organizations and individuals who master these skills will shape the future of work.
Learn Agent Architecture
Master the fundamentals of designing and implementing AI agent systems
Understand Governance
Learn how to create guardrails and safety mechanisms for agent deployment
Practice Collaboration
Develop skills for effective human-agent workflows and coordination
Build Responsibly
Embrace the principles of ethical AI development and deployment
Sign the Manifesto
Join technology leaders committed to responsible AI agent development
Add Your Signature
Current Signatories
Be the first to sign this manifesto