Category thesis

Agent work needs a production harness.

AI agents can write, inspect and repair software. What teams still lack is a production harness and control plane for deciding what counts as delivered. Orcho names that missing layer: typed plans, review contracts, readiness gates, event timelines and evidence bundles that survive the worker session.

PUBLIC ALPHA
Pipeline ⟳² ( ▶ plan [Claude]validate_plan [Codex] ) → implement [Claude]⟳² ( review_changes [Codex]repair_changes [Claude] ) → final_acceptance [Codex]
Agents
PLAN claude-opus-4-7 high
VALIDATE_PLAN gpt-5.5 medium
IMPLEMENT claude-opus-4-7 high
REVIEW_CHANGES gpt-5.5 medium
REPAIR_CHANGES claude-opus-4-7 high
FINAL_ACCEPTANCE gpt-5.5 low

01 Problem

Private sessions do not make a delivery system.

A chat transcript can explain what happened, but it rarely defines a durable plan, a gate result or a handoff decision. Teams need the run itself to carry the contract.

  • Plans should be typed enough for another phase to consume.
  • Review should alter control flow, not sit outside it.
  • Evidence should outlive the agent that produced it.

02 Protocol

The phases become the product surface.

plan, validate_plan, implement, review_changes, repair_changes and final_acceptance form the core delivery vocabulary. Each phase can emit artifacts and findings that later tools can inspect.

Next step

Move from the claim to the operating surface.

Inspect the technical contract, watch the proof, or run Orcho on one task. The deeper layer is always one explicit step away.