The AI copilot that's
actually yours.
You describe what you want in plain English. How the agent behaves — system prompt, model selection, tool policy, retries, when to pause, when to branch — is a file you wrote, not a binary we shipped. Edit it, version it in git, port it anywhere the engine runs. Any model, any vendor — or no vendor at all if you run your own.
Installs contenox. Run contenox init to scaffold a workspace.
What you can author
Glue work, with the policy on your side.
- →Someone yelled at you on Teams about a bug. Pipe the messages in; the chain checks the issue tracker for a duplicate and files it if not — pausing where you wrote it should pause.
- →It's Friday and you forgot the timesheet. Your git log knows what you did this week. The chain you wrote drafts the entries — your rounding rules, your project mapping, your approval gate.
- →New app on localhost:3000. You promised someone documentation. Playwright drives it; the chain decides what gets captured and where it lands. Notion, file, draft for review — whatever you authored.
- →Write the chain once, version it in git. Review each other's chains in PRs. The same artifact engineering already reviews; now your AI behavior lives there too.
In your editor
Drives Zed, JetBrains, and AionUi as an ACP agent.
Contenox speaks the Agent Client Protocolover stdio. Your chain runs inside Zed, a JetBrains IDE, or AionUi: tool calls render with the actual command, and HITL approvals route through the client's own permission UI.
// ~/.config/zed/settings.json
{
"agent_servers": {
"Contenox": {
"type": "custom",
"command": "contenox",
"args": ["acp"]
}
}
}We're onboarding design partners for Contenox Services and Contenox for Teams.
Each pilot is scoped individually — reach out and we'll figure out what makes sense for your team.
Request early access