prose

Multi-agent orchestration is managed by defining structured programs that specify agent interactions, data flow, and execution logic. This approach coordinates sequences, parallel tasks, and loops of AI agent sessions to automate complex processes.

Yes, parallel execution and context passing are supported natively. A coordinating agent can execute multiple specialized sessions concurrently, receive their outputs as context, and synthesize the results into a single actionable report.

State persistence supports multiple backends for flexible deployment, including file-system, in-context, SQLite, and PostgreSQL. This ensures workflow state is maintained reliably across different execution environments and scales.

Automating a code review process involves defining a declarative program where specialized agents (security, performance, style) review a pull request in parallel, followed by a coordinating agent that synthesizes their feedback into a single report.

No external framework is required. The declarative programming language natively manages loops, error handling, and state persistence within multi-agent workflows, enabling reliable execution of sophisticated AI-driven processes.

A declarative language is used when you need to define and execute complex multi-agent workflows with structured control flow. It suits scenarios requiring parallel execution, state management, and error handling across multiple AI agents.