First Run

Once you have an org, an agent, and a channel, you're ready to send your first signal and watch matic turn it into work. This section follows the full lifecycle of a single run: how an inbound signal becomes a task, how that task is decomposed and activated, how execution and review happen, and how the system records what it learned after delivery. Read these pages in order if you want a start-to-finish view of what matic does during its first complete cycle.

Create a Task

Create a Task explains how signals become tasks: sending a message, watching it get classified, and seeing matic produce a task spec with acceptance criteria.

Work Items & Work Pile

Work Items & Work Pile shows how a task decomposes into executable work items, each with a mandatory floor/ceiling work contract, and how they land on an agent's work pile.

Signals & Activation

Signals & Activation covers the routing pipeline that turns an inbound event into an activation through classification, interpretation, resolution, and the decision to begin work.

Run Lifecycle

Run Lifecycle describes what happens during a single run: context envelope assembly, agent runtime invocation, output production, and the immutable run record.

HITL Checkpoints

HITL Checkpoints explains where humans enter the loop through permission-gated approval gates, the dual-write rule for decisions and audit, and hard delivery blocks that require human override.

Review Artifacts

Review Artifacts covers validation of what the agent produced, including grounding checks, coverage checks against the work contract, regression checks against baselines, and previews for human review.

Observe, Hypothesize, Evolve

Observe, Hypothesize, Evolve explains the mandatory learning loop after delivery: registering observations, forming hypotheses from outcomes, and evolving the org through concrete changes.