That Principle of the least rights (Least Privilege) is a principle of authorization: A client only requests the permissions it needs for its intended operations. The Context Protocol (MCP) model embeds this principle in its authorization specification.
Implementation in the Model Context Protocol
In MCP, a client requests permissions in the form of scopes. A MCP server should specify in the WWW-Authenticate header which scopes are required to access a resource. The client thus immediately knows which scopes it should request. This way, it does not request excessive permissions. The scopes_supported field should indicate the minimum number of scopes for basic functions (Scope Minimization).
Right wing gradually expanding
The client gradually requests additional scopes through the so-called step-up authorization process. If a client lacks an authorization, the server responds with the HTTP 403 Forbidden error and the insufficient_scope message. The response lists the minimum scopes for the operation. The client then requests a new access token with extended scopes. To do this, it combines the previously requested scopes with the newly requested ones. After that, it repeats the original request, but no more than a few times. A client acting on behalf of a user should try this process. A client acting on its own behalf may try the process or immediately abort the request.
Similar approach for AI agents
When starting with AI agents, a similar approach applies. The agent initially only receives read-only permissions and is gradually given more authority. Sending, publishing, buying, deleting, and changing access rights remain the same for humans.
Demarcation
The principle of least rights limits what a client may do. Human-in-the-Loop and Human-on-the-Loop, on the other hand, regulate how a human monitors an AI system. Both approaches are related control principles, but not the same. Related entries: Model Context Protocol (MCP); Access Token (Access Token); Connector (AI); Authorization (SAP Business One); Human-in-the-Loop (AI).
AI in everyday work: Here's how to find the right start
AI assistant meets SAP Business One: Two paths out of the Versino AI shop
Sprint and Swirl: How an AI agent divides work and learns from trials
LATS: Considering several paths simultaneously with AI — and the availability check in SAP B1
ReAct and RLEF: How Artificial Intelligence Learns from Real Mistakes — and What Automated Bank Reconciliation in SAP Business One Has to Do with It
