MCP (Model Context Protocol)
An open protocol that standardises how AI models connect to external tools and data sources, so any compliant agent can use any compliant tool.
Before MCP, every AI product wired its integrations by hand: one bespoke connector per tool per vendor. The Model Context Protocol, introduced by Anthropic and since adopted broadly, defines a standard way for a model to discover a server's tools, read its resources and invoke its functions. For marketing stacks the consequence is practical: a platform that exposes an MCP server can be operated by whatever agent a team already runs, rather than only by the vendor's built-in assistant. It is becoming the test of whether a tool is genuinely agent-ready or merely has a chatbot attached.
Related terms
The mechanism by which an AI model invokes external functions, APIs or systems: the difference between an agent that can only talk and one that can act.
Coordinating multiple AI agents so their work composes: who runs when, what each is allowed to touch, and how their outputs feed each other.
An autonomous system that uses an LLM to plan, execute, and iterate on tasks with minimal human intervention.
Put these concepts into action
Oscom connects your SEO, content, ads, and analytics into one system. Stop context-switching between tools.
Start free trial