What Is Dry-Run Mode in an Ads MCP?
VigilDog Team · August 30, 2026 · 5 min read
Letting an AI assistant touch a live ad account is a reasonable thing to be nervous about. One misread instruction and a budget doubles or a campaign pauses. Dry-run mode is the design pattern that makes it safe: the assistant shows you exactly what it would change before anything actually happens. Here is what dry-run mode in an Ads MCP means in practice, and why it is the feature to insist on.
What dry-run mode means in an Ads MCP
An MCP (Model Context Protocol) server is the bridge that lets an assistant like Claude or ChatGPT call real tools, in this case, the Google, Meta, and LinkedIn ad APIs. A write tool such as 'create campaign' or 'update budget' would, by default, execute immediately against the live account. Dry-run mode inserts a stop before that happens.
In dry-run mode, when the assistant decides to make a change, the MCP does not call the ad platform to commit it. Instead it validates the request, resolves it against the real account (correct campaign IDs, valid budget values, policy checks), and returns a full description of what would happen: which entities change, from what to what, and any warnings. Nothing is written. You read the plan, and only if you approve does a second, explicit step carry it out.
How dry-run fits the request flow
The value is in the order of operations. The assistant proposes, the system previews, a human approves, and only then does the change reach the ad account. Reads, pulling performance, listing campaigns, checking search terms, flow straight through, because looking is safe. Writes are the ones that route through the preview-and-approve gate.
Dry-run vs approval gating
It helps to separate two ideas that often get bundled together. Dry-run is about simulation: computing and showing the effect of a change without committing it. Approval gating is about authorization: requiring a deliberate human 'yes' before a write executes. A well-built Ads MCP uses both, the dry-run produces the preview, and the approval gate ensures a person signed off on that specific preview before it becomes real.
The combination is what makes the difference between a tidy demo and something you would trust on a client account. A dry-run with no approval step still lets a change slip through automatically; an approval prompt with no real preview asks you to bless a change you cannot actually see. You want the preview to be accurate, resolved against the live account, not a guess, and the approval to be tied to that exact preview, so nothing changes between what you saw and what runs.
Why it matters for agencies
Agencies operate on other people's money and reputation, often across dozens of accounts. The failure modes of hands-off automation, a decimal in the wrong place, the right action on the wrong account, a pause that was meant as a test, are exactly the ones dry-run mode is built to catch, because you see the concrete plan before it executes. It turns 'I hope the assistant understood me' into 'here is precisely what it will do, approve or edit.'
It also creates a natural review trail. Because every write is previewed and approved, you have a record of what was proposed and what was sanctioned, useful for team oversight, for junior staff working under review, and for explaining to a client what happened and when. If you are new to driving ad accounts this way, our guide on how to run Google Ads from Claude shows the flow end to end.
Safety as the default, not a setting
The Ads MCP behind VigilDog is built on this exact model: reads flow freely, and every write is dry-run previewed and approval-gated across Google, Meta, and LinkedIn, so an assistant can do the tedious work while a human stays in control of anything that spends money. You get the speed of natural-language operations without handing over the keys. See what that covers and how teams use it on the pricing page, and if you manage client budgets, treat dry-run not as a nice-to-have but as the minimum bar.