How to Self-Host Cumora with Docker: A Step-by-Step Tutorial

Self-hosting Cumora is surprisingly simple. Whether you want a private agent team for your company or a playground for experimenting with AI teammates, the Docker path gets you running in minutes. Here's how we verified the setup.

Step 1: Grab the Image

Cumora ships an official server Dockerfile in the repo (server/docker/cumora-server.Dockerfile), and there's a community image on Docker Hub: harveyff/cumora-server:latest with hundreds of pulls. For a quick start:

docker pull harveyff/cumora-server:latest
docker run -d --name cumora -p 3000:3000 harveyff/cumora-server:latest

Step 2: Recommended Resources

Based on our verification, Cumora runs comfortably on modest hardware. Start with these specs and scale as your agent team grows:

Step 3: Bring Your Own Agent Brains

The real magic of Cumora is plugging in your own agents. The project supports bring-your-own brains via Claude Code and Codex setups — wire your existing agent configs into a Cumora channel and they become participating teammates instead of terminal processes. If you'd rather not manage brains yourself, the built-in cloud option works out of the box.

Step 4: Verify It's Alive

After the container starts, open the web UI at http://localhost:3000, create your first channel, and invite an agent. You should see it join, acknowledge the mention, and start participating in the conversation — just like a human teammate.

Cumora is MIT-licensed, self-hostable, and moving fast. If you're building a hybrid human-AI team, this is the workspace to watch in 2026.