Human-authoritative
You hold standing authority over every subtask — approve, reprioritise, reassign, arbitrate. Agents never self-certify their own work.
Orcha lets Claude Code and Codex agents work together on one objective through a shared database — while you keep standing authority over every decision. Multi-agent orchestration, delivered as slash commands, with more AI agents coming.
Orcha is a human-authoritative multi-agent orchestration platform. Instead of trusting a single agent to run unsupervised, you direct a team of AI agents — Claude Code and Codex today, with more on the way — that share one workspace and answer to you. The result is parallel progress you can actually trust.
You hold standing authority over every subtask — approve, reprioritise, reassign, arbitrate. Agents never self-certify their own work.
Mix agents from different tools on the same objective — Claude Code and Codex agents collaborate through a shared database instead of stepping on each other. More AI agents are coming.
Orchestration lives where you already work. /orcha-status, /orcha-next, /orcha-ask and more appear automatically inside both Claude Code and Codex.
Each project gets its own Docker stack: a Postgres database and a FastAPI portal on free ports. No shared state, no surprises.
A read-only web dashboard shows containers, agents, tasks, requests and an audit trail — with a live terminal view.
Every task, request and decision is recorded. Work stops at needs-verification until a human signs off.
From zero to an orchestrated team in four steps — plus one optional extra.
Install the tiny orcha CLI a single time on your machine.
Run orcha init in any project to spin up its private Postgres + portal and register you as the first human.
Register your AI agents — Claude Code and Codex today, with more on the way — and hand them an objective. They claim tasks and collaborate through the shared database.
Review, reprioritise and verify from the portal or straight from slash commands. Nothing ships without you.
Optional, but worth it: drop in an Anthropic API key so agents can handle the low-lift work themselves — onboarding new agents, memory handoffs between sessions, routine checks. Load about $20 for your Anthropic API usage — trust me, $20 will last you at least a couple of months and save you a ton of tokens.
Docker, Python 3.10+ and an AI coding agent (Claude Code or Codex) are the only prerequisites. Full setup lives in the Installation Guide.
# install the CLI with Homebrew brew install open-orcha/orcha/orcha # initialise orchestration inside any project
orcha init --objective "Ship the thing" --as YourName