MCP use cases

Start from the job rather than the tool. Each use case below is a working recipe: the task broken into steps, with the MCP servers that handle each one. 11 recipes across 5 roles.

Servers are picked from the catalogue and ranked by catalogue score, so a recipe names the best documented option for each step rather than every option. Where no server covers a step yet, the recipe says so.

Most popular use cases

The recipes people reach for first.

Marketing

Agents that research, write and run every part of a marketing motion.

Marketing work splits cleanly into research, production and reporting, and the MCP servers follow that split. The recipes below chain a data source to a writing tool to a place the output lands, which is the shape almost every marketing agent ends up taking.

Creative Production

Agents that generate and edit images, video, audio and music.

Creative work is the one area where an agent produces a file rather than text, which changes what matters. Check where each server writes its output, and expect to be billed per render rather than per request. Most of these wrap a paid generation API.

Ecommerce

Agents that run a storefront, its catalogue and its ad spend.

Store data and payment systems sit behind these servers, and several can modify live commerce data. Start read only, confirm the scope, then grant write access to the narrowest thing that works.

Engineering

Agents that read code, triage issues and touch infrastructure.

The best populated corner of the catalogue. These recipes assume an agent already has repository access and focus on what it does with it.

Operations

Agents that answer questions on your data and run internal workflows.

Operations recipes are mostly about joining a data store to a place people already look. The automation platform servers matter here, because they can trigger work that lives outside the agent entirely.