Everything new starts paused
Campaigns, ad sets, ad groups and ads are created paused. They don't serve and don't spend until you turn them on.
Connect Google Ads, Meta Ads and ChatGPT Ads once. Your AI assistant reads performance, finds wasted spend, drafts campaigns and makes changes — and every launch or budget increase waits for your approval.
Ad accounts I manage day to day — the same work Ads Connector now does from an AI chat.
Campaigns, ad sets, ad groups and ads are created paused. They don't serve and don't spend until you turn them on.
Launching, raising a budget or bid, publishing tags and deleting anything returns a preview first. The assistant has to show it to you and wait.
Each action the assistant takes is recorded with its arguments and result, so you can see exactly what changed and when.
Ad platforms to act on, analytics to check the result, and your own notes for context.
Search, Performance Max, Display, YouTube. Reports, search terms, keywords, budgets, bidding, new Search campaigns, and any GAQL query.
Facebook and Instagram campaigns, ad sets and ads. Insights by age, placement and device, creative fatigue, new image campaigns.
Chat-card ads inside ChatGPT answers. Campaigns, ad groups, creatives, location targeting and reporting.
Reports, realtime, key events — to see what ad clicks turned into.
Queries, pages, URL inspection, sitemaps.
Tracking audit, tags and triggers, publishing.
Paste one URL. The client opens a sign-in window, you approve access, and the tools appear.
claude mcp add --transport http ads https://ads-connector.com/mcp
[mcp_servers.ads] url = "https://ads-connector.com/mcp"
{
"mcpServers": {
"ads": {
"url": "https://ads-connector.com/mcp"
}
}
}{
"mcpServers": {
"ads": {
"httpUrl": "https://ads-connector.com/mcp"
}
}
}Any client that supports remote MCP servers with OAuth: Claude (web, desktop, Claude Code), OpenAI Codex, ChatGPT in developer mode, Cursor, Gemini CLI, VS Code. The model behind the client doesn't matter.
No. New items are created paused, and any action that can start or increase spend returns a preview until it's repeated with an explicit confirmation — which the assistant is instructed to request from you.
Refresh tokens and API keys are encrypted with AES-256-GCM and stored in a Cloudflare D1 database. They are only decrypted to call the platform on your behalf. You can disconnect any account at any time.
Reports pass through the server to your AI client and aren't stored. The server keeps your connections, default accounts, a log of changes and a monthly call counter.
The server can't reach files on your laptop, so the vault is read from the private GitHub repository the Obsidian Git plugin syncs it to, using a token limited to that one repository.
It's free during the beta.