Skip to main content

​
How MCP Works

The Model Context Protocol (MCP) is an open standard that lets AI assistants connect to external tools and data sources. Think of it as a universal adapter between AI and the real world.

​
The Problem MCP Solves

Without MCP, managing ads with AI requires:
  • Copy-pasting data between dashboards and chat windows
  • Building custom API integrations for each AI tool
  • No real-time data — AI works with stale information
With MCP, your AI assistant has direct, secure access to your ad platforms in real-time.

​
How Adspirer Uses MCP

  1. Your AI client (Claude, ChatGPT, Cursor, etc.) connects to the Adspirer MCP server
  2. The MCP server authenticates you via OAuth 2.1 and exposes 100+ tools
  3. When you ask your AI to do something (e.g., “create a campaign”), the AI calls the appropriate MCP tool
  4. The tool executes the action on the ad platform API and returns the result
  5. Your AI formats the response and shows you the outcome

​
Key Concepts

​
Tools

Discrete actions the AI can take — like create_search_campaign, get_campaign_performance, or research_keywords. Adspirer provides 100+ tools across 4 platforms.

​
Resources

Data the AI can read — like your business profile or strategy document. These provide context so the AI makes better decisions.

​
Prompts

Pre-built conversation starters that guide the AI through complex workflows like campaign creation or performance review.

​
Skills

Higher-level capabilities that combine multiple tools into cohesive workflows. For example, the “Campaign Management” skill chains keyword research → campaign creation → ad copy → extensions into a single guided process.

​
Security Model

  • OAuth 2.1 with PKCE — Industry-standard authentication
  • HTTPS/TLS — All data encrypted in transit
  • No data logging — Conversations and ad data are not stored by the MCP server
  • Scoped access — You control which ad accounts are accessible
  • Read-safe defaults — Destructive actions require explicit confirmation

Learn more about security

Read our full security documentation
Last modified on March 22, 2026