rune-orchestration

You can coordinate multi-agent workflows by partitioning tasks across language-model instances. This approach uses file-based handoffs and inscription-driven protocols to orchestrate, collect, and validate distributed outputs.

File-based handoff is a coordination pattern where agents write to and read from shared tmp directories to exchange findings. This enables asynchronous collaboration and output assembly across distributed language-model instances.

You can set up hierarchical coordination by defining teammate roles and validation pipelines through an inscription-driven workflow. This orchestrates specialized agents to concurrently review and audit complex codebases.

Peer-to-peer swarm coordination is structured using an inscription-driven protocol that defines roles and outputs. Agents operate autonomously and exchange findings via shared file directories to assemble a validated summary.

No, you do not need external dependencies to run multi-agent orchestration. The workflow relies entirely on partitioning tasks across language-model instances and using shared directories for file-based handoffs.

You should avoid distributed multi-agent orchestration for simple tasks that fit within a single context window. The overhead of managing file-based handoffs and inscription protocols is designed for complex code review and auditing workflows.