Hi HN, i am Dimitris, I have been using Claude Code and Codex agents, for some time now from the beggining i had been using them from inside my terminal mainly for coding. For the past 3 years i kept building so i have a homelab and a business server, which already has a lot of vms, so a lot of things to manage. So i decided to start building my own ideal version of using claude code, and codex flexibly and connect them easily with all my vms and infra but at the same time keep using them for coding, and as a personal assistant for daily use. So for the past months I built OtoDock. Its a self hosted platform, that runs real CLaude code and Codex as the engine. On top of them i build a live dashboard with websockets and integrated in depth the clis, so its even nicer to work with the cli through the dashboard instead of working on the terminal. I even ended up integrating interactive terminals as well (due to the subscription change Anthropic was trying to do for harnesses) so you can also use interactive terminals directly from the dashboard remotely. OtoDock is a multitenant self hosted agents platform, is meant for businesses to create and manager agents, that can act autonomously on the background doing various jobs, or they can also be collaborative, they can work with users (team members) of the platform so many team members can use a collaborative agent meant for a specific job or department (ex. social media agent, dev agent, etc.) Each user of the platform can bring his own anthropic and openai subscriptions that is automatically used for his user sessions with the clis. (or local models through Ollama) I have also built an MCP framework, so that its easy to add cummunity mcps fast on the application, and i keep a community mcp repo where i will put all available mcps there. The dashboard as well has a lot of features, for example it has a fully functional chat for talking to the agents, that supports collabora previews for editing and viewing PDF, Word, Excel, Powerpoint files as the user works with the agents, a workspace per user with a proper file browser (like each user and agents google drive showing in the dashboard), per agent shared knowledge and workspace folders across users, a per user and per agent memory system, scheduled tasks with real intervals, webhook triggers that can initiate a scheduled task of an agent, meetings between agents (live meetings directly in the chat where the agents can collaborate), voice with TTS and STT for talking to the agents and many more... I am a security paranoid myself so i spent a lot of time making sure the agents run as securely as possible and can be safely desployed in a real business server. The agents run on a kernel sandbox with network isolation by default, and specific local up excemptions are added by the admin for specific mcps that need local network access. Access to the internet of course is always on. The license, it is Fair Source, not OSI open source. Self hosting is free up to 5 users, all the code is public so you can review everything. Trying it is one docker compose file, images are on GHCR, the setup wizard creates your admin locally, no signup or email anywhere. Its the first time I publish a proper application and I would love people to try it and read the code and use it in their setups, I have spent a lot of time on this and i would love all the feedback I can get. AI Disclaimer: Large parts of OtoDock are written using otodock itself, for the last 2 months the application was mature enought where i started using an agent of OtoDock daily for the development. If you use Claude Code Cli or Codex for development feel free to try it, you might find very useful and helpful in your workflow. (remote machines feature will be available in a week, where you will be able to make the agents of the platform run in your local machine). GitHub: https://github.com/OtoDock/oto-dock Website: https://otodock.io |