Platform
One delivery protocol around many agent workers.
Orcho is the stable layer around agent work: typed plans, phase gates, review loops, handoff decisions, evidence and run history. Start here when you want the product model before going into docs.
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.
/platform/agent-delivery-protocol/Not every task deserves the same workflow depth.
Orcho separates two decisions that agent tools often blur together. A work-kind profile says what the task is — small_task, feature, complex_feature, planning, review, research or migration. An operating mode says how much verification and evidence the run should require — fast, pro or an explicit governed override.
/platform/control-depth/Your agents will change. Your delivery protocol should not.
Orcho ships runtimes for Claude Code, Codex, Gemini and Claude-compatible wrappers, with a registry for plugin runtimes. Workers remain replaceable participants under one delivery record, so changing a worker does not erase plan history, gates or evidence.
/platform/workflow-portability/Top-model intelligence belongs where it matters most.
AI coding spend grows when every phase uses the same expensive worker. Orcho makes runtime choice explicit: plan with strong reasoning, route bounded implementation and repair through cost-efficient compatible workers, keep review independent, and inspect the result by phase.
/platform/runtime-economics/Review should control the run.
Orcho puts gates and human handoff inside the delivery flow. A rejection requests retry feedback, routes to repair or halts the run — and the decision itself becomes part of the run history.
/platform/evidence-and-governance/Start with one task
See the protocol work before reading every layer.
Install the public alpha for a self-serve run, or open the proof hub and inspect why a delivery was approved or blocked.