prose

You orchestrate multi-agent AI workflows by writing declarative programs in the OpenProse language to define sequences, parallel tasks, loops, and error handling. This approach uses a VM-like execution model to manage task orchestration, dynamic context passing, and state persistence.

The best way to automate a multi-step AI process is to define reusable agents with specific models and prompts in a .prose file. You can compose these agents into sophisticated systems where one identifies issues, another suggests fixes, and a third synthesizes feedback.

Declarative syntax handles parallel execution and error handling by allowing you to explicitly define execution paths and fallbacks in structured programs. The VM-like execution model manages state persistence and dynamic context passing between specialized agents.

No specific dependencies are required to run recursive program composition for AI agents. The system manages agent specialization, dynamic context passing, and recursive composition internally through its VM-like execution model without external dependencies.

Yes, you can use structured programs instead of simple chat prompts to create robust, repeatable AI processes. By writing human-readable programs in .prose files, you move beyond simple chat interactions to define sophisticated multi-agent systems with state persistence.

Multi-agent task orchestration fails without state persistence because the execution model relies on maintaining state across parallel tasks and recursive program composition. Structured programs need state management to correctly pass dynamic context between specialized agents during execution.