Skip to main content
Adspirer sits between your AI assistant and your ad platforms. You talk to your AI in natural language. Adspirer translates that into authenticated API calls to Google Ads, Meta Ads, Amazon Ads, ChatGPT Ads, LinkedIn Ads, and TikTok Ads — and returns results back to your conversation.

The Full Pipeline

Every interaction follows this path. Whether you’re pulling a performance report or creating a campaign, the pipeline is the same.

Layer 1: Your AI Client

Your AI assistant — Claude, ChatGPT, Gemini CLI, Perplexity, Cursor, Codex, or any MCP-compatible client — is the interface. You describe what you want in plain English:
  • “How are my Google Ads performing this month?”
  • “Create a LinkedIn campaign targeting IT Directors”
  • “Find wasted spend across all platforms”
The AI determines which Adspirer tools to call and in what order. With skills installed, it follows proven workflows. Without skills, it makes its best guess.

What the AI Client Provides


Layer 2: MCP (Model Context Protocol)

is the protocol that connects your AI client to Adspirer. It’s an open standard — any AI client that supports MCP can use Adspirer.

What MCP Does

  • Tool discovery: Your AI client discovers Adspirer’s 400+ tools with descriptions and parameters
  • Tool invocation: When the AI decides to use a tool, it sends a structured request via MCP
  • Response delivery: Adspirer returns structured results that the AI formats for you
  • Authentication: OAuth flows are triggered transparently through MCP

One URL, All Platforms

Every AI client connects to the same MCP endpoint:
This single URL gives access to all 400+ tools across all 6 ad platforms. Your AI client doesn’t need to know about Google Ads APIs, Meta APIs, or LinkedIn APIs — Adspirer handles all of that behind the MCP layer.

Tool Categories

Adspirer exposes tools in two categories:
  • Read tools pull data and analyze performance. Safe to run anytime. No spending impact. Examples: get_campaign_performance, research_keywords, analyze_wasted_spend.
  • Write tools create or modify campaigns. Require user confirmation. Affect spend. Examples: create_search_campaign, update_campaign, add_keywords.

Layer 3: Authentication (OAuth 2.1 with PKCE)

When you first connect an ad platform, Adspirer uses OAuth 2.1 with PKCE — the same standard used by banking apps.

The OAuth Flow

1

You trigger a connection

Either through your AI client’s MCP settings or by running setup (/adspirer:setup).
2

Browser opens

You’re redirected to the ad platform’s login page (Google, Meta, Amazon, ChatGPT, LinkedIn, or TikTok). You sign in with your existing credentials.
3

You authorize permissions

You see exactly what Adspirer is requesting (read campaigns, create ads, manage budgets). You approve.
4

Tokens are issued

The ad platform issues an access token (1-hour lifetime) and refresh token (30-day lifetime). Adspirer stores these encrypted at rest.
5

Automatic refresh

When the access token expires, Adspirer automatically uses the refresh token to get a new one — no re-login needed until the refresh token expires (30 days).

What Adspirer Never Sees

  • Your ad platform password
  • Your billing/payment information
  • Data from accounts you haven’t connected
  • Personal data beyond what’s in your ad account settings

Layer 4: Ad Platform APIs

Adspirer translates MCP tool calls into native API calls for each platform: Each API call uses the OAuth token from your authenticated session. Adspirer handles rate limiting, pagination, error handling, and response formatting.
Marketing stack integrations (Google Analytics, Klaviyo) are brokered through Pipedream Connect — the OAuth token lives in Pipedream scoped to your Adspirer user, and Adspirer proxies tool calls to Pipedream’s remote MCP. Adspirer never sees the raw third-party credentials. Available on Plus, Pro, and Max plans.

Layer 5: Skills & Agent (Optional)

Without skills, your AI calls tools individually and guesses the workflow. With skills and the agent, you get structured workflows, brand awareness, and strategy persistence.

What the Agent Adds

See Performance Marketing Agent for the full architecture.

Safety Model

Adspirer’s safety model operates at multiple layers: No single layer is sufficient on its own. The safety model works because each layer reinforces the others.

Data Flow Summary

Adspirer reads your campaign data in real-time and passes it to your AI client. It does not permanently store campaign data, ad copy, or targeting settings.
Last modified on July 8, 2026