MCP — Marketing Tools in the Mag.AI-Marketing Curriculum
Castalia Institute — Mag.AI-Marketing
Authoritative copy: part of the MyST project (myst.yml).
What MCP is (here)¶
The Model Context Protocol (MCP) is an open standard for connecting AI assistants to external tools and data through well-defined servers. In this program, marketing tools means any capability you would use in real go-to-market work — analytics, CRM, ad platforms, content systems, search/social APIs, internal spreadsheets — exposed as MCP servers (or consumed from approved community servers) so workflows stay inspectable, scoped, and reproducible.
We care about MCP because it forces explicit contracts: what the model can do, with what credentials, and what gets logged — instead of opaque “the AI did something in the browser.”
Curriculum alignment: Mag.AI-Marketing names major ecosystems students model and may connect: Meta (Facebook/Instagram), Google (Ads, Analytics/GA4-style, Search/YouTube), Amazon (marketplace + Sponsored Products/Brands + retail media)—see CURRICULUM_FULL.md (Major platform ecosystems). MCP servers might wrap read-only reporting pulls, creative asset lists, or synthetic datasets with the same schema shape as production exports.
Build vs leverage¶
| Mode | What students do |
|---|---|
| Leverage | Call existing MCP tools from notebooks or agents: pull metrics, draft from templates, query segments — always with documented assumptions and evaluation |
| Build | Implement or extend MCP servers that wrap marketing APIs, data stores, or simulators (including PTAH/iNQspace hooks): define resources, tools, and prompts with least privilege |
Courses expect both: you must use tools well before you ship tools others depend on.
How terms use MCP¶
Term I (Audience & Message): Instrument personal and audience models with live signals where ethical (e.g. your own site analytics, public trend APIs) via MCP; design tool schemas that map to ontology choices.
Term II (Campaign & Growth): Orchestrate multi-touch workflows; attribute and experiment with data pulled through MCP, not screenshots.
Term III (Strategy): Portfolio and automation courses require governed MCP surfaces (audit trails, escalation) aligned with AINS-M6205.
Specific course artifacts may require an MCP appendix: tools used, servers touched, and failure modes.
Guardrails (non-negotiable)¶
Consent & ToS — Only connect accounts and data you are allowed to use for coursework; respect platform policies.
Secrets — No API keys in Git; use environment / secret stores as required by iNQspace or faculty policy.
Scope — Tools are narrowly described; avoid “full admin” MCP surfaces for class demos.
Measurement — When MCP feeds a model, document latency, bias, and missing data as you would for any sensor.
Access and equity¶
Not every student has ad accounts, spend, or employer API access. Faculty may provide:
Shared or faculty-hosted MCP endpoints with read-only or synthetic data
Paired projects where one student supplies account access under policy and another builds analysis only
Fully synthetic endpoints that still exercise tool contracts and logging
Course artifacts must be achievable without self-funding ad spend; live spend is optional enhancement, not a hidden prerequisite.
References (external)¶
Legal¶
“The Castalia Institute Magisterium confers proprietary credentials based on demonstrated work and evaluation. These credentials are not accredited academic degrees and do not confer professional licensure.”