DAM Butler MCP
Natural-language retrieval across 235,000+ global brand assets, with an architecture adopted and shipped into daily enterprise workflows in Breville Group global GTM teams.
The problem
Breville’s Brandfolder DAM holds 235,000+ digital assets: product imagery, brand materials, and marketing content across global markets. GTM teams in APAC, North America, and EMEA needed those assets daily.
Getting to the right file meant knowing the exact folder structure, taxonomy, and metadata conventions. Brand managers, marketers, and content producers who didn’t already know the system had one option: ask someone who did. That created a repeatable bottleneck at exactly the point where speed matters.
What was built
Breville’s first Model Context Protocol (MCP) tool.
A custom GPT connects to Brandfolder through two sequential stages. First, an intent parser converts any natural-language query into a structured API request, mapping descriptions, markets, dates, and product names to the right filter parameters. Then a clarification loop runs for ambiguous inputs before the API call fires, asking one targeted follow-up rather than returning a wall of partial matches.
The result: ask “the white Barista Express hero shot approved for EU markets, updated after January 2025” and get back the right asset with metadata, directly in chat. The user never touches the DAM interface or learns the taxonomy.
Result
Prototyped September 2025, demoed to Breville product leadership, and the architecture was then adopted and shipped to production by the BRG engineering team.
Now in daily workflows for GTM teams across APAC, North America, and EMEA.