Ad Growth
Reads your ad account every morning, pauses losers and scales winners on written, versioned rules, and asks before any budget goes up. It turns Ad Creative's batch into a managed account.
Who it is for
The owner who checks Ads Manager at midnight and changes budgets on feel, and the performance marketer who wants every change logged with the rule that fired.
What it reads from the brain
Before every run. Sections are addressed as layer.section.
What it does, step by step
- Read the brain
Opens the sections above and the inputs before touching the account.
- Gather evidence
Lists campaigns and pulls insights through the read tools. Every finding cites its source.
- Decide and draft
Writes the kill and scale rules for your account and a ranked plan. Drafts use your customers' vocabulary and only claims_allowed.
- Advise stops here
You get the account read, the rules and the plan.
- Act
Applies rule fired pauses directly, because pausing never costs money. Requests approval for every budget increase, every activation and every new ad set.
- Write back
Learnings appended through the brain writer: what worked, what did not, the decisions log.
Tools it uses
meta_ads.campaigns.listWhat is running and how it is structuredmeta_ads.insightsDaily performance per campaign, ad set and admeta_ads.ad.pauseRule fired pauses. The one action that runs without a tap, because it spends nothingmeta_ads.budget.setApproval gated, alwaysmeta_ads.campaign.activateApproval gated, alwaysmeta_ads.adset.createApproval gated, always. Audience tests launch as paused drafts
The three modes for this agent
Read tools only. You get the account read, the written kill and scale rules, and a ranked plan. New customers stay here for the first 14 days while the rules are tuned.
Pauses that a rule fired happen directly. Every budget increase, activation and new ad set pauses for your approval.
Daily at 07:00. Same gates. Nothing on the money or messaging list runs without a tap.
Guardrails and approvals
Enforced in code. Items with a filled circle wait for your tap in Telegram.
- Total daily spend cap never exceeded.
- No single change over 20 percent of a budget per day.
- Never creates a campaign without approval.
- Every change logged with the rule that fired.
- New customers stay in Advise for 14 days while rules are tuned.
- Kill and scale rules are written and versioned in the brain.
- Every recommendation cites its evidence. Uncited is removed.
- The summary states what data level it ran at and what a higher level would add, once.
What it writes back to the brain
Dated and sourced. Learnings and performance sections are append only.
Example output
Example output. Structure only. No performance figures.
Ad Growth, your store, daily read
Ran at data level 3 (Meta connected). Under 250 words.
What each rule did today
- Kill rule v2, condition met on 2 ad sets: paused. Evidence: meta_ads.insights, last 3 days, ad set ids listed in changes.csv.
- Scale rule v2, condition met on 1 ad set: budget increase drafted, capped at 20 percent. Waiting for your approval.
- Audience test: 1 new ad set drafted paused. Waiting for your approval.
Not changed: everything else. Daily cap untouched.
rules.md (versioned in the brain)
- v2, dated: Kill when an ad set meets the written spend and result condition over the written window. Scale when it meets the written efficiency condition for the written number of consecutive days, by no more than 20 percent per day.
- v1, dated: superseded, kept for the log.
Also delivered: account-read.md (what each campaign did and which rule fired), changes.csv (every budget or status change with reason).
Planned: the exact rule thresholds shown in this example are set per account during the 14 day Advise period and are not published here.