Start with the question you already have
You open ChatGPT and ask: “Who is still waiting for a reply from me this week?” ChatGPT alone does not know.
When AlefOS is connected through MCP, the assistant can query the work memory your account is authorized to access. You can ask: “What did I promise this customer?”, “Why do I need to call them back?”, “What changed in the last meeting?” or “What should I know before tomorrow’s appointment?”
What MCP changes
MCP is the structured connection between the assistant and AlefOS. The assistant does not receive unrestricted database access. AlefOS remains responsible for scope, permissions, available tools and controlled actions.
ChatGPT, Claude and other supported assistants are conversational interfaces. AlefOS is the memory and permission boundary behind the answer.
From an answer to a prepared follow-up
You ask: “Prepare the follow-up for this customer.” When the relevant tool and plan allow it, the assistant can use authorized context to prepare the next step. Protected execution remains controlled by AlefOS.
The useful change is not that the assistant guesses an answer. It is that it can work from the customer history your account is permitted to use.
FAQ
Does MCP expose the whole AlefOS database?
No. AlefOS applies authentication, workspace scope, plan, role, permissions and handler-level rules before returning context or allowing an action.
Is AlefOS only a ChatGPT integration?
No. AlefOS is the work-memory and action layer. Supported assistants provide interfaces to that authorized context.
Does the assistant execute everything automatically?
No. AlefOS can prepare supported next steps, while protected actions remain subject to the confirmation behavior required by the relevant workflow.
Where are the connection instructions?
Use the official MCP Cockpit documentation and the current public endpoints generated from AlefOS runtime sources.
Explore AlefOS
Continue through the product documentation, core modules, integrations and plans.