prose

You orchestrate multiple AI agents by writing structured programs using a declarative syntax that supports sequential execution, parallel processing, loops, conditional branching, and error handling. This approach allows you to define complex agent behavior and control flow within a single program.

To manage state persistence across AI agents, you can use file-based, SQLite, or PostgreSQL storage backends. This ensures context is reliably passed and maintained between agents throughout the workflow execution.

Yes, you can create reusable agent definitions and facilitate modularity through program composition. This allows you to build complex multi-agent workflows by combining smaller, reusable program components.

You automate a multi-step code review by defining a structured program where one agent analyzes code, another checks for security vulnerabilities, and a third synthesizes findings into a report. The declarative syntax coordinates these agents sequentially or in parallel.

Yes, this AI orchestration approach supports conditional branching and error handling directly within the structured programs. You can define complex control flow patterns to manage agent execution paths and manage exceptions gracefully.

To start programming AI agent workflows, you need to write structured programs using the declarative syntax. You can begin by running the example program 'examples/01-hello-world.prose' to understand the basic structure and execution.