Plain-language programming for building AI agents using Markdown + YAML operation blocks
Fractalic is a plain-language programming environment for building AI workflows and agents using ordinary Markdown plus lightweight YAML operation blocks. Instead of wiring nodes in a UI or writing imperative Python scripts, you express intent in structured text.
Key idea: A Fractalic file = (Ordered Knowledge Blocks) + (Operation Blocks that transform context) → Final Returned Result.
# Product Requirements {id=reqs}
- Fast response times
- Easy to use interface
- Reliable performance
@llm
prompt: "Create a technical specification based on these requirements"
blocks: reqs
use-header: "# Technical Specification"
mode: appendAfter execution, your document grows:
# Product Requirements {id=reqs}
- Fast response times
- Easy to use interface
- Reliable performance
@llm
prompt: "Create a technical specification based on these requirements"
blocks: reqs
use-header: "# Technical Specification"
mode: append
# Technical Specification {id=technical-specification}
## Performance Requirements
- Response time: < 200ms for all user interactions
- System uptime: 99.9% availability target
- Concurrent users: Support up to 10,000 simultaneous connections
## User Interface Specifications
- Responsive design for mobile and desktop
- Intuitive navigation with maximum 3-click access to any feature
- Accessibility compliance (WCAG 2.1 AA standards)
## Reliability Standards
- Automated failover mechanisms
- Data backup and recovery procedures
- Comprehensive error handling and logging The document evolves as operations run—new blocks are appended, replaced, or refined—so you can literally "grow" an AI system the way you draft a document.
Ready to build your first AI workflow? Check out the Quick Start guide or dive into Core Concepts to understand the fundamentals.