Tools
106 tools across four ad platforms. Every tool is one of three flavors:
- Read (most tools): Pulls data. Always safe to call.
- Transform: Reshapes or summarizes data without hitting the platform.
- Write: Changes something (pause, edit budget, add negative keywords). Off by default. Turn on per provider in app.adsyncs.net/settings.
You don't memorize tool names. The AI does. Browse by platform to know what's possible.
Browse by platform
Google AdsSearch, Performance Max, YouTube: pull metrics, plan keywords, and audit account health.
Meta AdsFacebook + Instagram campaigns: read campaigns, ad sets, ads, breakdown reports, audiences.
LinkedIn AdsB2B campaign management: manage campaign groups, target by company or job, batch updates.
Bing Ads (Microsoft Advertising)Microsoft Search Network: keywords, search terms, Performance Max, LinkedIn-profile targeting.
How the AI finds the right tool
When you ask a question, the AI picks tools by what they do, not what they're called. So "show me last week's worst-performing Google campaigns" turns into a call to google_ads_get_campaign_performance automatically. You never type the name.
If you ever want the AI to explore on its own, there's a meta-tool:
ads_discover_tools(query: "find tools for keyword research")
It returns a ranked list of tools that match. Useful when you're prototyping a new workflow and want to see what's available without scrolling this page.
Naming convention
Every tool is prefixed with its platform: google_ads_*, meta_ads_*, linkedin_ads_*, bing_ads_*. That means even with all four providers connected, you can scan the tool list at a glance.
Every tool carries _meta["ads/tier"] (read / transform / mutation), _meta["ads/category"], and _meta["ads/tags"]. MCP clients can filter on these to build power-user UX. The ads_discover_tools meta-tool indexes the same fields.
When something goes wrong
Sometimes a tool can't finish: an ad platform is having a bad day or sends back something unexpected. When the problem is on our side (not a login or a typo you can fix yourself), the AI will offer to flag it for us:
"That failed on Adsyncs' end. Want me to report it to the team?"
Say yes and the AI files the report for you: what it was trying to do, what it already tried, and its read on what went wrong. No copying error text into a form, no leaving your chat. It's asked every time, so nothing leaves your workspace unless you agree, and the AI is told never to include logins or secrets, with anything sensitive stripped out before it's stored. Right after sending, the AI shows you exactly what was filed, so there are no surprises about what left your chat.
This is how we hear about real problems fast, from the one place that actually saw them happen.
What about prompts and resources?
Prompts (pre-baked workflows like "weekly account audit") and resources (read-only context like "workspace connections") live on Resources & prompts.