Understanding MCP, The 'Universal Adapter' For AI In Advertising | AdExchanger

"Metaphorically speaking, it's plumbing for AI: the piping that lets large language models connect to other software in a predictable and controlled way. In practical terms, it's a specification, which was developed by Anthropic and open-sourced in late 2024, for how models interact with other software so AI can perform tasks across different systems, like pulling data and kicking off processes."

"Developers define which actions their software can perform using the MCP standard format, including what data each action touches. AI systems can then string those actions together - for example, 'pull a report,' 'create an audience,' 'send a message' and then 'launch a campaign' - to complete a workflow."

"AdCP, WebCP and UCP all use MCP as their underlying transport, but each one speaks to a different set of jobs. AdCP is for advertising workflows, such as campaign setup and bidding. WebCP is for web actions and navigation. And UCP is for commerce use cases, all the way from product discovery to checkout."

Model Context Protocol (MCP) is a specification developed by Anthropic and open-sourced in late 2024 that functions as foundational infrastructure for AI systems to interact with other software. It establishes standardized rules for how models access tools, call functions, and operate within defined limits. Developers define actions their software can perform using MCP's standard format, specifying which data each action touches. AI systems can then combine these actions sequentially to complete workflows. Industry-specific protocols like AdCP for advertising, WebCP for web navigation, and UCP for commerce are being built on MCP's foundation, eliminating the need for custom integrations and allowing all to plug into the same agent ecosystem.

Read at AdExchanger