Collaboration

Agent collaboration in matic is a set of named, protocol-driven primitives — not implicit coordination. Every handoff, delegation, consultation, and conflict carries defined semantics, ownership rules, and a recorded outcome. This section covers how agents coordinate work, transfer context, resolve disagreements, and manage parallel execution within and across teams.

Agent Messages

The structured signal types agents use to communicate with each other — message schemas, routing rules, and how agent messages move through the Signal pipeline.

Collab Modes

The four named interaction patterns (solo, pair, swarm, supervised) that govern how agents share a workspace, coordinate writes, and transition authority during execution.

Conflict Resolution

How matic detects and resolves incompatibilities between agent outputs, competing hypotheses, asset merge conflicts, and policy disputes — including escalation paths and the role of Auto Matic as arbitrator.

Delegate and Consult

Two distinct collaboration actions for multi-agent work: delegate assigns a sub-task to another agent while the delegator retains ownership; consult requests focused input without any transfer of authority.

Handoff

Full ownership transfer of a Work Item from one agent to another — including what gets packaged, how the receiver acknowledges, and what the sender's state becomes after transfer.

Helpers

Ephemeral, anonymous execution units spawned within a parent agent's run — what Helpers can and cannot do, how their outputs are attributed, and why they have no promotion path to persistent agents.

Work Pile and Dispatch

How agents manage their ordered queue of pending work, how the Dispatcher identifies independent Work Items for parallel execution, and how Budget Gates and Bandwidth Probes constrain concurrency.