MCP overview
What the Model Context Protocol is and how MyAd exposes its ad tools.
The Model Context Protocol (MCP) is an open standard that lets AI assistants call external tools. MyAd runs an MCP server over your connected accounts: it reads live Google Ads and Meta Ads data and, with a write-scoped key, pauses and enables campaigns, adds keywords and negatives, changes budgets and bids, and applies health recommendations.
Why it matters
Instead of switching dashboards, you ask your AI assistant:
“Audit my biggest account, list the wasted keywords and add the obvious negatives.”
The assistant looks up the catalogue with tools/list, calls a tool with your accountId, and gets back the same
numbers and recommendations the dashboard shows.
Endpoint
POST https://myad.si/api/v1/mcp
The server speaks JSON-RPC 2.0 and implements:
initializetools/listtools/callping
Access
Requests carry an API key created in Dashboard → Settings → API, scoped to specific accounts and to read-only or read-write. A read-only key can call every read tool; a write tool needs the write scope. OAuth clients can request the same scopes.
The tools
The server publishes read and write tools for accounts, health checks and recommendations, performance reports,
keywords and search terms, budgets, bids and assets, and the changes themselves.
Browse the full directory at Tools reference; the API serves the same catalogue at GET /api/v1/tools.
Writes need approval
Read tools never change anything. Changes that cannot add spend run at once; anything that can add spend becomes a pending request the account owner approves or rejects within 48 hours, and approval re-reads the account before it applies. Campaigns MyAd’s autopilot (托管) manages are refused. Tools reference has the details.
Safety
- Every tool respects the key’s scope and account list.
- Write tools are logged to History with the request, the result and who decided.
- Spend-raising changes need an extra confirmation, and one request can at most double a budget or bid.