Glossary
Ads MCP glossary: MCP, approval gates, ROAS, MER, and attribution terms
Last reviewed: October 20, 2026 · AdsCove Product & Engineering
Plain definitions for the terms you meet when you connect an AI agent to Google Ads, Meta Ads, TikTok Ads, and Shopify. Each entry says what the term means in general and, where it matters, how AdsCove implements it, with a link to the page that documents the detail.
MCP and connection terms
- MCP (Model Context Protocol)
- An open protocol that lets an AI client call tools and read data from external systems through an MCP server, instead of copying data into the chat by hand. Learn more: MCP for ecommerce agents →
- MCP server
- The service that exposes tools to the AI client. An ads MCP server exposes tools such as campaign reports or budget changes for connected ad accounts. Learn more: AdsCove product docs →
- Remote MCP endpoint
- An MCP server reached over HTTPS rather than run on your computer. AdsCove uses one remote endpoint for every supported platform: https://adscove.com/api/mcp. Learn more: Supported platforms →
- MCP client
- The AI app that connects to the server, for example Claude, ChatGPT, Cursor, Codex, Gemini, or Windsurf. Each client has its own place to add a server URL. Learn more: AI client setup guides →
- Custom connector
- Claude's name for a remote MCP server you add yourself by pasting its URL in Settings → Connectors. Learn more: Connect AdsCove to Claude →
- OAuth vs API key
- Two ways a client proves who you are. AdsCove supports OAuth for clients that support remote-MCP OAuth, and an Authorization: Bearer API key for clients that do not. Learn more: Security model →
- Progressive tool catalog
- Keeping the default tools/list small and letting the agent discover more tools on demand. In AdsCove the agent calls discover_adscove_tools, then run_adscove_tool with the tool name. Learn more: Live tools reference →
Write-safety terms
- Read tool vs write tool
- A read tool only returns data (reports, settings). A write tool changes something in the account (budget, status, a new campaign). Reads can run directly; writes deserve review. Learn more: AI ad operations safety →
- Approval gate
- A rule that a proposed write waits for an explicit human decision before it runs. In AdsCove a protected write returns a pending execution tied to an execution_id; nothing changes until it is approved with approve_action, and reject_action discards it. Learn more: Approval-gated ad changes →
- Dry run
- Planning a change without submitting it, so you can read the exact payload first. A dry run never mutates the account. Learn more: Approve or reject agent changes →
- Idempotency key
- A unique key attached to a write so that sending the same request twice does not execute it twice. It protects against duplicate budget changes or duplicate campaigns when an agent retries. Learn more: Approve or reject agent changes →
- Paused draft (PAUSED / DISABLE)
- A new campaign or ad created switched off, so it cannot spend until someone activates it separately. AdsCove's Meta and Google creation tools create PAUSED resources; TikTok creation tools use DISABLE. Learn more: Meta Ads MCP →
- Audit history
- A record of every tool call: what was asked, what ran, and the result. AdsCove keeps redacted or size-limited inputs and outputs with risk level, and before-and-after snapshots for supported writes. Learn more: Security model →
Measurement terms
- Platform-reported ROAS
- Return on ad spend as each ad platform reports it: the platform's attributed conversion value divided by spend. Each platform uses its own attribution settings, so totals can overlap. Learn more: Meta ROAS vs Shopify →
- Blended ROAS
- Store revenue divided by total ad spend across platforms. AdsCove's blended_roas is Shopify gross order value ÷ combined ad spend for the same date range. Learn more: ROAS methodology →
- MER (marketing efficiency ratio)
- Total revenue divided by total marketing spend, without trying to credit individual campaigns. Because AdsCove's blended ROAS uses all Shopify revenue, it behaves like MER rather than campaign-level ROAS. Learn more: ROAS methodology →
- Attribution gap
- The difference between what ad platforms report and what the store recorded. AdsCove computes it as platform-reported revenue − Shopify gross order value; a positive gap is a diagnostic signal, not proof of a specific tracking error. Learn more: Weekly diagnosis: attribution gap →
- UTM parameters
- Tags added to a landing URL (utm_source, utm_medium, utm_campaign) that tell the store where a visit came from. AdsCove totals orders with first-visit UTMs as supporting context next to blended ROAS. Learn more: Meta ROAS vs Shopify →
- Shopify as source of truth
- Treating the store's recorded orders as the final word on revenue and using pixel-reported revenue only as a reference when deciding budgets. Learn more: Shopify connection →
- MCC (Google Ads manager account)
- A Google Ads account that manages other Google Ads accounts. AdsCove supports MCC access and asks the agent to confirm the customer ID before analysis. Learn more: Google Ads MCP →
Glossary FAQ
What does MCP mean in advertising?
MCP (Model Context Protocol) is how an AI client such as Claude or ChatGPT calls tools on an ad platform through an MCP server: reading campaign reports, or preparing changes such as budgets and new campaigns.
What is the difference between ROAS and MER?
ROAS credits revenue to ad spend, usually per platform or campaign using that platform's attribution. MER divides total revenue by total marketing spend with no campaign credit. AdsCove's blended ROAS uses all Shopify revenue, so it is closer to MER.
What is an approval-gated write?
A change the agent proposes that does not run until a person approves it. In AdsCove, protected writes return a pending execution; approve_action runs it and reject_action discards it.