Model Context Protocol (MCP)
Open protocol · Connecting AI agents to tools and data SIGNET support: Architectural alignment — forward-looking
What it is
The Model Context Protocol is an open protocol for connecting AI models and agents to tools, data sources, and services through a standard interface. Originated by Anthropic and now broadly adopted across the AI ecosystem, it has become the common way an agent discovers and invokes capabilities exposed by the systems around it.
How SIGNET supports it
MCP sits in SIGNET's architectural alignment tier, with the qualifier stated plainly: the current specification defines no normative MCP binding. The alignment is one of design, and it is forward-looking.
SIGNET's Agent Layer is transport-agnostic by construction. An AgentCapability describes what an agent may do independently of how it is invoked; a Mandate bounds the agent's authority regardless of the protocol carrying its actions; and every material act produces a Decision with provenance, whatever interface triggered it. The consequence is that the governance travels with the agent, not the transport: a SIGNET-governed agent whose capabilities are exposed as MCP tools remains bounded by its mandate and accountable through its decisions, exactly as it would over any other interface.
That makes the natural relationship complementary rather than overlapping. MCP standardises how an agent reaches tools; SIGNET standardises what a commerce agent is authorised to do and how its actions are recorded. A normative binding — advertising AgentCapability objects over MCP tool interfaces, and binding MCP-invoked actions to mandates and decisions — is a candidate for a SIGNET extension, to be taken through the change-proposal process rather than asserted ahead of it.
Where it lives in the standard
- Agent Layer — SyntheticAgent, AgentCapability, Mandate, Decision
- Extensions — how a binding would be proposed