Everyone’s “doing AI” these days. Many products or platforms try to solve genuine problems, while others just add the magic button with stars graphics and hope people won’t ask what it actually does.
In short, there are good and bad ways of implementing AI into products, and with Voluum MCP we believe we found the perfect one.
Why We Chose MCP?
AI agents seem like a natural fit for marketers who use Voluum. AI is great in finding patterns in large pools of data, something that Voluum users are used to do. But there are some serious caveats to ANY method of implementing AI, and we’ve considered and weighted them all.
Automatic AI optimization
We could built on the already existing and successful Traffic Distribution AI feature and create some kind of auto-optimization mechanism. Just as reminder: Traffic Distribution uses machine learning to direct traffic to most profitable pages. It is a very good tool for this specific task and we could have this on a much bigger scale.
The users would get the magic “Optimize with AI” button, the marketing would basically write itself.
But we knew that adding placements, 30+ data points and custom variables, campaign budgets and assumptions, and all other aspects of running a campaign to the mix, would render this optimization… well, not optimal. How would you even know that this optimization worked, that it actually is delivering the best results. How would you verify what was done to your campaigns? Would you even trust this button?
We believe our users deserve full control over what happens with their data and money.
Built-in AI agent
The other popular route is to create an embedded AI agent. You talk with it, give instructions, and it fulfils your instructions. This is a convenient solution, as it puts everything inside one platform, however, it comes with some serious asterisks:
- You don’t know how capable the model is. Is it on par with the lates advances in AI development? Will it be updated regularly?
- You don’t have control over your data. Do you trust the model to give it full access to your data? Wouldn’t you prefer to give it read-only access?
- You can’t connect other platforms. Cross-platform integration is something that makes AI agents truly shine, and you cannot connect anything external to built-in bot.
This was our reasoning when we selected to go with MCP.
What is MCP?
MCP, or Mobile Context Protocol, is a connectivity standard for AI agents, first developed by Anthropic, the makers of Claude. Think of it as USB for AI agents: it allows them to connect to a wide variety of tools:
- Communicators
- Data analysis tools
- CRMs
- Ad networks
MCP gives users versatility and control. It is YOU who selects the AI engine; it is YOU who defines the type of access an agent should have, it is YOU who adds more integrations to connect the whole marketing stack.
The Benefits of Using Voluum MCP
We’ve decided to go the MCP route, empowering our users to use AI as they seem fit. It wasn’t the easiest decision: MCP is harder to communicate than bult-in copilots or AI auto-optimization. But it is the right way.
- Voluum MCP allows you to select an AI agent you already use and trust. You probably already have an agent that knows how you work, and the only thing it was lacking is access to your data.
- Connecting your agent to Voluum allows you to run scheduled tasks. From morning digests to alerts, from performance reports to clients or supervisors to opportunity finders, AI agent can work not only when you ask it but also in the background
- Having an external AI agent allows you to precisely define the scope of access to your data. Give read-only access or enable write options in a clearly defined scope.
To sum up, MCP is the only solution that really allows you to work from one prompt:
- Create campaigns in ad networks and Voluum simultaneously.
- Send notifications to a communicator based on data gathered by Voluum
- Pull data from CRMs and analytics software like Voluum.
- Optimize all marketing efforts at once: adjust your budget setting and tune traffic targeting s in ad networks, re-route traffic in Voluum, suggest new angles and opportunities.
This is the way to go.
How to Make The Most Out of your MCP-cconnected AI agent?
1. Bulk domain and campaign-URL updates
Problem. When you migrate to a new tracking domain, or a redirect domain gets flagged, updating the URL on dozens or hundreds of campaigns by hand is slow and easy to get wrong.
What the agent does. Ask the agent to swap the domain across every affected campaign in a single instruction. It shows you the list of what will change, then applies the update in bulk once you confirm.
SAMPLE PROMPT
Change the tracking domain on every campaign using go.olddomain.com to go.newdomain.com. Show me the full list before applying anything.
2. Bulk create and update offers and landers
Problem. Building or editing large batches of offers and landers is tedious, and the manual upload view limits how many rows you can add at once.
What the agent does. The agent creates or updates offers and landers in bulk from a file or a plain-language description, past the manual row limits, and places them in the workspace you choose.
SAMPLE PROMPT
Create offers from this CSV, mapping column A to the name and column B to the offer URL, and put them all in the Nutra workspace.
3. Bulk notes and tags
Problem. Keeping notes and tags consistent across many campaigns, offers, and landers is manual housekeeping that rarely gets done.
What the agent does. The agent applies or edits notes and tags across any set of entities you describe, so your organization stays consistent without click-by-click effort.
SAMPLE PROMPT
Add the tag Q3-push to every campaign created this month and set their note to “Reviewed by media team”.
4. Account cleanup: archive, restore, and move
Problem. Old and inactive campaigns clutter your account, and archiving, restoring, or moving them between workspaces one at a time is a chore.
What the agent does. The agent archives, restores, or moves entities in bulk based on the criteria you set, so cleanups that used to take an afternoon happen in one instruction.
SAMPLE PROMPT
Archive every campaign with zero visits in the last 90 days, and move all campaigns tagged client-A into the Client A workspace.
5. Custom multi-dimensional reports
Problem. The standard report view limits how many dimensions you can group by at once. And because reports only return entities that received traffic, offers and paths that got no visits never appear, so you cannot easily see what is sitting idle inside a campaign.
What the agent does. The agent pulls reports grouped across as many dimensions as you need and exports them to CSV. To reveal what got no traffic, it reads the campaign setup, lists every offer and path configured in it, and cross-references that against the report, so idle offers and paths surface even though the report on its own would leave them out.
SAMPLE PROMPT
Show me the last 30 days for campaign X grouped by country, device, and offer. Then list any offers configured in the campaign that received no traffic, and export both to CSV.
6. Cost-discrepancy investigation
Problem. Tracked cost, ad-network spend, and reported conversions do not always line up, and tracking down where they diverge is slow detective work.
What the agent does. The agent cross-checks tracked cost against ad-network spend and conversions, then tells you exactly which campaigns diverge and by how much.
SAMPLE PROMPT
Compare tracked cost against the ad-network spend for my Facebook campaigns last week, and list the ones that differ by more than 5%.
7. Cost updates
Problem. Keeping campaign costs accurate requires regular manual updates, including per custom variable or per offer, and even for periods with no visits.
What the agent does. The agent updates costs on the schedule and rules you define, covering manual cost, cost per variable, and cost per offer, including where there were no visits.
SAMPLE PROMPT
Set the cost for campaign X to 0.02 per click for all of yesterday’s visits, including where there were no conversions.
8. Total spend and conversion caps with auto-pause
Problem. Built-in caps are daily. There is no lifetime or total cap that automatically pauses a campaign once a spend or conversion ceiling is reached.
What the agent does. The agent watches your campaigns and pauses them automatically the moment a total spend or total conversion limit is hit, filling a gap the daily caps leave open.
SAMPLE PROMPT
Watch campaign X and pause it as soon as total spend reaches 5,000 or total conversions reach 200. Check every 15 minutes.
9. Scheduled start, pause, and dayparting
Problem. You want campaigns to run only at certain hours, or to pause when performance drops, without watching the dashboard around the clock.
What the agent does. The agent starts, pauses, and adjusts bids on a schedule or when your conditions are met, so time-of-day and day-of-week logic runs itself.
SAMPLE PROMPT
Pause my push campaigns every night from 1am to 6am UTC and resume them each morning.
10. Traffic-distribution and weight optimization
Problem. Splitting traffic optimally across offers and paths is hard to tune by hand, especially with many paths or a goal the built-in optimizer does not target.
What the agent does. The agent computes optimized path and offer weights from recent performance and applies them, and it can optimize toward your chosen goal such as ROI, CPA, or CPM.
SAMPLE PROMPT
Rebalance the path weights in campaign X to maximize ROI based on the last 7 days, and apply the new weights.
11. Domain and offer-URL health monitoring
Problem. A domain that goes down, a certificate that expires, or a flagged offer URL can quietly burn budget until you happen to notice.
What the agent does. The agent checks your domains and offer URLs on a schedule, alerts you to problems, and can reroute or pause the affected campaigns automatically.
SAMPLE PROMPT
Check all my custom domains every hour. Alert me if any go down or a certificate is expiring, and pause campaigns using a dead domain.
12. Error-log monitoring and triage
Problem. The error log is noisy and easy to ignore until something is already broken.
What the agent does. The agent reviews the error log, groups issues by type, and gives you a short prioritized summary of what actually needs attention.
SAMPLE PROMPT
Summarize my errors from the last 24 hours by category, and tell me which ones are likely hurting conversions.
13. Conversion uploads and custom-conversion management
Problem. Uploading conversions by CSV is fiddly, and managing custom-conversion data in bulk is manual.
What the agent does. The agent uploads conversions and manages custom-conversion data for you, straight from a file or a description of the mapping.
SAMPLE PROMPT
Upload the conversions in this CSV, matching click IDs in column A with payouts in column B.
14. Scheduled reporting and custom digests
Problem. You want a regular performance summary delivered to you, beyond the fixed views the dashboard offers.
What the agent does. The agent builds the exact report you want and delivers it on a schedule, to Slack or email, highlighting whatever matters most to you.
SAMPLE PROMPT
Every Monday at 8am, send me last week’s spend, revenue, and ROI by campaign, and flag anything down more than 20%.