<p align="center">
  <img src="docs/images/logo/agenvo-pigeon.png" alt="Agenvo logo: a coral-colored homing pigeon in flight." width="160" height="160">
</p>

<h1 align="center">Agenvo</h1>

<p align="center">Let your AI assistant coordinate coding agents on your own computers and servers.</p>

<p align="center"><a href="README.zh-CN.md">简体中文</a></p>

Fork of [Xuanwo/agenvo](https://github.com/Xuanwo/agenvo) (Apache-2.0) adding the ZCode connector, a dark relay admin theme and headless credential issuing.

Ask ChatGPT on your phone to check on Codex running on your server, or have an agent in Herdr on your laptop fix a failing test. Agenvo connects your assistant to [Herdr](https://herdr.dev) and [Codex app-server](https://developers.openai.com/codex/app-server/) through MCP, using a relay you host on Cloudflare or a single VPS. You describe the goal; your assistant decides how to move the work forward.

Experimental [Amp support](docs/amp.md) connects native Amp hosts through a local plugin. An experimental [ZCode Connector](docs/zcode.md) attaches the local ZCode CLI's app-server the same way.

![Agenvo architecture. The assistant layer shows ChatGPT, dots, Grok Bot and Muse. Your assistant sends MCP calls over HTTPS to the Agenvo relay you host on Cloudflare or a VPS, and the relay sends change events back. On your laptop and server, the Herdr and Codex connectors connect out to the relay over WSS; requests and results travel over that connection. The pigeon identifies Agenvo components. Each connector passes work to Herdr or Codex app-server, where agents run with your projects and local tools. Your computers need no inbound ports.](docs/images/architecture.png)

## In practice

- **Leave your desk, keep the work going.** Your assistant sees the agents already running in Herdr without restarting them, and can resume Codex threads later. Check progress or reply from your phone.
- **Follow up when something changes.** Ask a [ChatGPT dot](https://learn.chatgpt.com/docs/dots) to watch a computer's agents and tell it how to respond. When an agent finishes a turn, waits for input or exits, the dot receives an event, reads the output and takes the next step.
- **One assistant across your computers.** Herdr on your laptop and Codex on your server appear in one list. Your assistant can start new work where it fits: a Codex thread on the server, or Codex, Claude or Devin in a Herdr pane.
- **Agents keep going through a connection loss.** Closing the chat or losing the network does not stop them. Connectors reconnect on their own, and your assistant picks up from each agent's current state.

## Set up

Ask your coding agent to set up Agenvo with the [installation guide](docs/installation.md).

## If you are an AI agent

- Set up Agenvo for the user and connect their computers and assistant: [Installation](docs/installation.md).
- Inspect a formal release notice and update an existing deployment: [Updating](docs/updating.md).
- Manage agent work through an existing MCP connection: [Agent management](docs/management.md).
- Watch for agent changes and react to them: [Runtime events](docs/events.md).
- Diagnose connection, authorization, or failed calls: [Diagnostics](docs/usage.md#revoke-and-diagnose) and [logs](docs/logging.md).
- Understand or change Agenvo's architecture: [Architecture design (Chinese)](design/architecture.zh-CN.md).

## License

[Apache-2.0](LICENSE).