Platform · AI agents
AI agents need a role, just like every new colleague.
In roleALPHA an AI agent gets its place in the organisation: a role, the workflow it contributes to and the rules that apply to that role.
Role·Remit·Accountable
Workable by people
Sound familiar?
- Every new agent gets its rules copied into a prompt by hand — and each one a slightly different version.
- A responsibility changes. The people know, the agents keep working on the old state.
- Data protection or the works council ask which agent was allowed to know what. There is no answer.
What changes
- 01 A role instead of one-off instructions An agent fills a role, supports workflows and is subject to that role's policies — from the same source as the people.
- 02 Changes apply to both When a responsibility changes, the agent retrieves the new version the next time. Nobody has to keep a second version in step.
- 03 Accountable Which agent retrieved which context, and when, is logged — for audit, data protection and works councils.
Runnable by agents
Context instead of programming
An agent retrieves its frame through an open interface (Model Context Protocol): its role, its responsibilities, the rules that apply. Its access is a key of its own, revocable at any time, and allows exactly this: reading its own context.
The operating system
Part of one model
An agent is part of the same model as the people: it fills roles, works in value streams and brings competences that can be described.
FAQ
Frequently asked
Does roleALPHA run the agents?
No. Your agents run where they run today. roleALPHA gives them a role, a remit and rules, and logs which context was retrieved.
Can an agent change data through roleALPHA?
The agent access can only read its own context. It has no write permission and sees no data outside its bundle.
Which AI providers work?
Any agent that speaks the Model Context Protocol. For the built-in assistant you choose provider and model yourself; it is switched off by default.
What if an agent's rules are wrong?
Then you correct the role or policy — as a proposal with approval, like any other change. The agent gets the new version on its next retrieval.
See it on your own organisation.
One conversation is enough to see whether this fits your organisation.
Workable by people. Runnable by agents.