prose

Multi-agent workflow orchestration coordinates parallel tasks, agent definitions, and context binding to maintain durable run state across AI sessions using filesystem, in-context, SQLite, or PostgreSQL backends for persistence and resumption.

Stateful execution for parallel AI agent tasks is managed by orchestrating session spawning and parallel branches, binding results to drive downstream tasks, and persisting run state to database or filesystem backends for reliable resumption.

OpenProse VM supports multiple state backends for workflow persistence, including filesystem, in-context, SQLite, and PostgreSQL, allowing you to resume complex multi-agent execution states across different storage environments.

To pass context between agents in a multi-agent workflow, you use context binding semantics to attach and pass execution results, enabling downstream tasks to consume the outputs of upstream agent branches and imports.

You can run a simple .prose file to execute a basic multi-agent workflow session, allowing the OpenProse runtime to orchestrate session spawning, agent imports, and context passing automatically.

You should avoid multi-agent orchestration for AI workflows when your tasks are strictly sequential, require no parallel branches, or lack the need for durable run-state management and context binding across persistent sessions.