Campaign Manager 360 MCP Server
Programmatic
Connect your AI assistant to Campaign Manager 360 through one secure MCP link — analyse delivery, recover wasted spend and pace budgets, with every change held for human approval.
What is the Campaign Manager 360 MCP?
The Campaign Manager 360 MCP Server connects your AI assistant directly to Google’s Campaign Manager 360 (CM360) through OpenAds — a single, secure Model Context Protocol connection. Instead of clicking through Trafficking, Reporting and Attribution interfaces, you brief an assistant like ChatGPT, Claude, Claude Code, Cursor or Codex in plain English. It reads your advertisers, campaigns, placements and creatives, pulls performance data, and drafts the changes you ask for — while OpenAds acts as the control layer that keeps analysis, proposal, approval and execution strictly separate.
This matters because CM360 sits at the centre of enterprise ad serving and measurement, where a single misconfigured placement or tracking tag can distort attribution across every channel. OpenAds lets the AI do the heavy analytical lifting — reconciling impressions, spotting pacing gaps, surfacing wasted spend — but never lets it execute a write action silently. Every campaign edit, placement change or creative assignment is proposed first and held for a human to approve, so you gain speed without surrendering governance.
Key capabilities
Read advertiser, campaign, placement and creative structures across CM360 accounts.
Pull delivery and performance metrics — impressions, clicks, conversions, viewability and cost.
Reconcile pacing against flight dates and booked budgets to flag under- or over-delivery.
Draft new campaigns, placements and creative assignments for review.
Identify wasted spend across low-viewability or non-converting placements.
Generate Floodlight and conversion reports with attribution context.
Propose optimisation and budget-pacing changes with clear before-and-after logic.
Hold all write actions for explicit human approval before execution.
What you can do
Analyse performance
Ask the assistant to summarise how any advertiser or campaign is delivering across a date range. It reconciles impressions, clicks, conversions and viewability by placement and site, highlights the strongest and weakest performers, and explains movements against benchmarks — turning raw CM360 delivery data into a clear read on what is working.
Research keywords & audiences
While CM360 is a serving and measurement platform rather than a keyword tool, the assistant analyses which audience segments, sites and placements convert best. It cross-references Floodlight activity and audience lists so you can prioritise the placements and segments that drive real outcomes and retire those that only spend.
Find wasted spend
The assistant scans placements for low viewability, high invalid-traffic signals, zero-conversion delivery and duplicated tags. It ranks the biggest sources of leakage in ₹ terms and proposes pausing or reallocating them — so you recover budget that would otherwise serve impressions no one values.
Create campaigns
Describe an objective, flight dates and budget, and the assistant drafts a campaign with structured placements and creative assignments mapped to the correct advertiser. Nothing is trafficked live; the full structure is presented for your review and only built in CM360 once you approve.
Optimise & pace budget
The assistant compares actual delivery against booked budgets and flight timelines, then proposes pacing adjustments — accelerating under-delivering placements or throttling those burning budget too fast. Each recommendation shows the current pace, the target and the expected impact before you decide.
Generate reports
Ask for a weekly delivery report, a Floodlight conversion breakdown or an advertiser-level summary, and the assistant assembles it with the metrics and attribution context you specify — ready to share with stakeholders in seconds rather than hours.
Supported campaign types
Campaign typePrimary useTypical workflow Standard displayAwareness and reach across the open webDraft placements, assign creatives, pace to flight dates VideoIn-stream and outstream brand storytellingSet viewability and completion goals, monitor delivery Programmatic (via DV360 linkage)Automated, audience-driven buyingReconcile delivery, optimise placement mix Rich media / interactiveEngagement-led creative formatsAssign creatives, track interaction events Tracking-onlyThird-party measurement and attributionDeploy Floodlight tags, verify conversion capture RetargetingRe-engaging known audiencesMap audience lists to placements, monitor frequency
Getting started
Create or sign in to your OpenAds account and open the integrations directory.
Select the Campaign Manager 360 MCP Server and start the connection flow.
Authorise OpenAds against your Google Marketing Platform account with the appropriate CM360 network and advertiser access.
Choose which advertisers and profiles the assistant may read and propose changes for.
Confirm your approval settings so that every write action requires explicit sign-off.
Once connected, link your AI assistant of choice — ChatGPT, Claude, Claude Code, Cursor or Codex — to the same OpenAds MCP endpoint. The assistant inherits your permissions and approval rules automatically, so you can begin briefing it in natural language from whichever tool you already work in.
First prompts to try
"Summarise delivery for the Summer Sale advertiser over the last 14 days."
"Which placements have viewability below 50% but more than ₹50,000 in spend?"
"Show me the top five converting placements by Floodlight activity this month."
"Draft a display campaign for a ₹8,00,000 budget running 1–31 August."
"Where are we over-pacing against booked budget right now?"
"Build a weekly conversion report grouped by site and creative."
"Find duplicated or non-firing tracking tags across the Retail advertiser."
"Propose reallocating ₹2,00,000 from underperforming placements to our best performers."
A complete workflow example
Objective: recover wasted spend on a display advertiser mid-flight.
User prompt: "Our Diwali display campaign has a ₹12,00,000 budget and finishes in ten days. Find wasted spend and propose fixes."
What OpenAds analysed: the assistant read every placement in the campaign, pulled impressions, viewability, clicks and Floodlight conversions, and reconciled current pace against remaining budget and flight dates.
Findings: three placements had consumed ₹2,80,000 with viewability under 40% and zero conversions, while two high-viewability placements were under-delivering and set to leave ₹1,50,000 unspent.
Proposed actions: pause the three low-value placements, reallocate their remaining budget to the two strong performers, and lift daily pacing on those to close the flight fully.
Approval step: OpenAds presented the changes as a proposal with the ₹ impact of each. Nothing was trafficked. The user reviewed, adjusted one reallocation, and approved.
Result: only on approval did OpenAds execute the edits in CM360. The campaign redirected ₹2,80,000 of wasted budget toward converting placements and finished the flight fully paced — with a complete audit trail of what changed and why.
Best practices
Brief the assistant with specific advertisers, flight dates and budgets so its analysis stays precise.
Review proposed placement and tag changes against your measurement setup before approving.
Use viewability and conversion thresholds to define wasted spend consistently across advertisers.
Schedule regular pacing checks mid-flight rather than waiting until budgets are exhausted.
Keep approval checkpoints on for all write actions, even for routine optimisations.
Safety and control
OpenAds draws a hard line between read and write actions. Reading data — pulling metrics, reconciling pacing, drafting reports — happens freely so the assistant can analyse without friction. Every write action, however, from pausing a placement to trafficking a new campaign or editing a tracking tag, is held at an approval checkpoint. The assistant proposes; a human approves; only then does OpenAds execute. This separation means the AI can move quickly on analysis while you retain full authority over anything that touches live delivery, with an audit trail on every change.
OpenAds vs manual Campaign Manager 360 management
DimensionOpenAdsManual CM360 Analysis speedNatural-language queries answered in secondsManual report builds and exports Wasted-spend detectionProactive scans ranked in ₹ termsAd hoc, dependent on analyst time Change controlEvery write action held for approvalDirect edits, reliant on process discipline Audit trailAutomatic record of proposal and approvalManual change logs, if kept Reporting effortAssembled on request in plain EnglishBuilt manually in the reporting interface
Limitations
CM360 is a serving and measurement platform, so keyword-level bidding lives in linked buying tools rather than here.
Some programmatic buying actions require Display & Video 360 and depend on the linkage between platforms.
The assistant works within the advertiser and profile permissions you grant — it cannot see or change accounts outside that scope.
Attribution insights reflect your Floodlight and conversion configuration; misconfigured tags limit what can be analysed.
Frequently asked questions
Does OpenAds change my campaigns automatically?
No. The assistant only analyses and proposes. Every write action is held at an approval checkpoint and executes solely after a human approves it.
Which AI assistants can I use?
Any assistant that supports MCP, including ChatGPT, Claude, Claude Code, Cursor and Codex. They all connect to the same OpenAds endpoint and inherit your permissions.
Can it manage programmatic buying in DV360?
Some programmatic actions require Display & Video 360 and the linkage between it and CM360. OpenAds reconciles and reports on that delivery within the access you grant.
How does OpenAds handle permissions?
You choose which advertisers and profiles the assistant can read and propose changes for. It cannot act outside that scope, and approval rules apply to every write action.
Will it interfere with my Floodlight tracking?
The assistant can propose tag changes, but nothing is deployed without approval. Your existing measurement setup stays intact unless you explicitly sign off on a change.
Connect it through one secure MCP connection.