Each example below is a prompt you can try into your AI assistant once the MCP is connected, adapted with your own campaigns, time periods, or workspaces. The AI answers with your real numbers, in English or French depending on the prompt's language.
Explore your data
The MCP queries raw data in your workspaces directly, on any scope: Ads, campaigns, keywords, site analytics...
Ad performance
"Compare CTR across my META ads over the last 30 days"
"Which 5 creatives have the best ROAS this week ?"
"Top 10 keywords by clicks this week on Google Ads"
By campaign or channel
"Compare ROAS between channels last month"
"Which campaigns are not on track on their budget?"
“What's my spend this month, detailed by channels ?”
Breakdowns (country, age, gender, device, placement...)
"What's my ROAS by country this month?"
"Breakdown impressions by age on my latest Meta campaign"
"Mobile or desktop: where do I convert best?"
Site (Analytics)
"How many sessions generated by Linkedin Ads last week?"
“What’s my revenue on Google Ads this month?”
"What's my bounce rate by channel?" be careful to use your dimension name if your analytics is matched. Otherwise, you can ask data per standard analytics dimensions (UTM source…)
Track your goals and pacing
If goals (ROAS, CPA, budget...) are set up on your workspace, the AI can tell you where you stand without opening spyne.
"How are my goals doing this month ?"
"Am I on track on my ROAS goal ?"
"Which goal is furthest behind right now ?"
"Sum up where I stand on each goal in one plain-language sentence."
“In workspace [workspace ID], which campaigns are over- or under-pacing this month broken down by [channel]?”
Check your data freshness
Before disputing a number or heading into a call, ask the AI to confirm your data is current.
"My Meta numbers look off, is the connection up to date?"
"When was my Google Ads connection last synced?"
"Is any integration erroring out on this workspace?"
"List my active integrations on this workspace."
Manage your workspaces
For multi-workspace accounts (agencies, multi-market setups), the AI can list and cross-reference your workspaces to help you keep a clean view of your account.
"List all my spyne workspaces."
"Which workspaces have no active integration connected?" ; candidates to clean up or archive.
Automate your recurring tasks
You’ve got a bilan to fill in at the end of every campaign, every week, every month and feeding it with the latest numbers each time takes time. Our clients already use spyne to automate the creation of these bilans: they connect their agent to both spyne (via MCP) and their existing template, and have it refill the template with the current numbers. Same layout, same sections, every time.
Monthly client bilan
“Using the attached template, fill in [Client]’s bilan for [Month] with data from workspace [ID], spend, ROAS, and top campaigns by platform.”
“Prepare last month’s PowerPoint bilan for [Client] from workspace [ID]: spend, impressions, reach and CPM by channel and by campaign, with subtotals by channel and a grand total.”
How totals work and their limits: Totals and subtotals with spynemcp
Weekly numbers for a presentation
“Pull this week’s spend, ROAS, and conversions for [workspace/client], in the same format as last week, so I can drop them straight into the deck.”
Tip: keep the template and the request itself consistent from one run to the next, same structure, same dimension/metric names, that’s what keeps the numbers comparable over time.
Coming soon
Auditing your custom dimensions
"Analyse my custom dimensions on this workspace, and which ones have a lot of unmatched rows?"
"Propose a rule to create a 'Product' dimension from my Meta campaign names, and test it"
"Help me find custom dimension ideas regarding my campaign taxonomy"
You can’t write in spyne, but you can test and analyze your dimensions.
Auditing your analytics matching
”Get the active Analytics Matching configuration for workspace [ID] and test every rule for 01-08-2026 to 31-08-2026. Give me a table with the overall matching rate (sessions and clicks) for the source "Google Ads", and flag anything that looks broken or worth investigating.”