AI News Feed
Market watch
Products & Applications

Developer Releases OtoDock, a Self-Hosted 'Company OS' Running Claude Code and Codex Agents

Developer Dimitris has published OtoDock, a Fair Source, self-hostable company OS that runs Claude Code and Codex CLI agents as sandboxed persistent processes and works with Anthropic or OpenAI subscriptions or local models.

OtoDock is multi-tenant by design, so several people can collaborate on company agents, according to the post. The developer says it supports four different modes of collaboration and runs on an Anthropic or OpenAI subscription, or with local models.

Each agent runs on Claude Code or Codex CLI as a persistent process on the user's server inside a bubblewrap kernel sandbox, with network isolation enabled at all times through pasta, according to the post. Every agent is given its own workspace, memory, schedules and tools. The same agents can also be configured to run identically on any remote computer through a single outbound WebSocket connection, which the developer says requires no inbound ports and no VPN.

The post lists a set of built-in capabilities: agents can answer and place phone calls through Twilio or a user's own Asterisk system, and can edit video as well as Excel, Word and PowerPoint files, with those files previewed directly inside the chat through Collabora. The developer says more features are included beyond these examples.

Installation is described as one install script and a Docker Compose file, with no signup required. OtoDock is published under a Fair Source license, all of the code is public, and self-hosting is free for up to five users, according to the post. The project's source is hosted at github.com/OtoDock/oto-dock, with a website at otodock.io.

In a disclaimer, the developer states that large parts of OtoDock are written using OtoDock itself, running Claude Code. He also asks readers to review the code and examine the sandbox model, writing: "I would love people to read the code, take a look in the sandbox model, and tell me what would stop you from running this on your own hardware."

At the time described in the post, the Hacker News submission showed 3 points and no comments.