Hubzoid
A self-hostable platform for internal enterprise AI agents that keeps a team's knowledge, instructions, skills and tools in a versionable folder and reuses that shared context through chat, scheduled workflows, or a personal assistant connected over MCP.
Hubzoid gives a team's internal AI agents a single shared hub of context, skills and tools, defined in markdown so it can be versioned like code. A user can start a useful agent alone on a laptop, then share it with a team through accounts and scoped permissions, keeping business data in its source systems and connecting it through tools or MCP rather than copying it in.
A chat interface answers questions and takes authorized actions using the hub's shared context; a workflow engine built on DBOS runs scheduled markdown tasks and Python workflows, producing inspectable reports; and hub tools, knowledge and skills can be brought into a separately running personal AI assistant through MCP. A minimal starter template ships a sample skill, knowledge file, custom tool and sub-agent to inspect and adapt.
Built with Python 3.11 or 3.12, installed with pip and run with a hubzoid CLI; the workflow engine needs SQLite 3.42 or newer, or PostgreSQL. Defaults to local single-user mode, with authentication available for shared deployments. Apache-2.0 licensed.