OtoDock Launches Self-Hosted Agent Platform Built on Claude Code and Codex

A newly released self-hosted platform called OtoDock acts as a central brain for organisations, enabling teams to deploy autonomous agents that operate in departments and collaborate with each other. The software is multi-tenant by design and runs on a Linux server requiring Docker, a minimum of 4 GB of RAM, and a script to handle the initial deployment stack.

Engineers can configure each agent to run on different AI engines, choosing between individual user subscriptions for Claude Code and Codex, or local models operating directly on internal hardware. The platform allows agents to execute tasks on a schedule, respond to webhooks, or trigger once at a chosen time while communicating with internal tools via the Model Context Protocol.

Security is handled through kernel-level sandboxing with constant network isolation, requiring explicit service-by-service permission grants for every action. Administrators can also pair external workstations running macOS, Linux, or Windows via a single outbound connection that maintains file synchronisation without needing open ports on the local network.

The platform features a setup wizard accessible at localhost:8400 that establishes an owner account and installs a personal assistant equipped with necessary tooling during the initial run. Users can extend functionality by building custom agents or importing them from a community catalog, while managing access levels through four distinct sharing and operational modes.

Licensing follows a fair source model under the Functional Source License, v1.1, with an Apache 2.0 future grant that automatically applies two years after release. Self-hosting remains free for up to five users, whereas growing teams transition to seat-based licensing using signed keys without any feature gates placed on underlying data.

Why it matters for AI for Software Development

For software developers managing multiple AI workflows, OtoDock provides a structured method to run Claude Code and Codex processes inside isolated sandboxes on local infrastructure. It bridges terminal-based AI tools with company services, letting engineers delegate routine tasks to agents while maintaining strict control over API subscriptions, workspace files, and network security boundaries.

What to do about it

• Review the system requirements including Docker and a minimum of 4 GB of RAM on a Linux server.
• Set up the environment by executing the installation script within a designated folder.
• Connect personal Claude or ChatGPT subscriptions under the setup interface to enable the AI engines.

Tools mentioned

• ChatGPT
• Claude
• Claude Code

Source: Show HN: Self-hosted company OS, Claude Code and Codex agents in departments (github.com)