Release context
Current hosted tasks prepare research, briefs and copy. Finished image rendering is separate. The current licensed edition excludes the separate performance-audit and insights service.
This is a dated release record. Historical model names, rates, trial limits and capability counts do not set today’s offer. Current task inputs, source coverage and your review govern use. Ad Library active status and longevity do not establish spend, profitability, a test result or why an ad stopped. A 5×5×5 copy set is preparation, not 125 completed experiments.
What to inspect first.
| Your next step | Work and review |
|---|---|
Inspect | Choose the historical mode that matches the question, then check its current supported counterpart. |
Retain if you try it | Keep that task’s required inputs, text/file output and missing-evidence notes together. |
Review before use | A dated mode list does not establish current account operations or finished rendering; inspect today’s task and cost boundary. |
Source: Meta Ads Agent v1.1.0 · dated published release · Reading guidance from the published release; no exercise, generated artifact or customer result is claimed here.
Use the instructions and compatibility guidance for the version and access you actually have. Review the current supported option →
What changed
Open the concise change summary
9 modes
Research to launch plan - the full system surface, hosted
Video scenarios, roasted
Spine, beats, persona-locked VO, storyboard - adversarial panel verdict included
Live Ad Library spy
Coverage maps, white space, and faithful template rip-off briefs
7-step diagnosis
AD vs FUNNEL vs OFFER vs TARGETS, with max 3 routed actions
Release notes
One day after launch, the Meta Ads Agent graduates from a creative-foundation preview to the full hosted surface of the production system.
New modes
- Big-Idea pipeline - ideation fan-out (8-12 UMP/UMS candidates per awareness tier), hard-rule exclusions, scored survivors, hybridization of the top 3, and customer-voice expansion of the winner.
- Static batch - 5 concepts per run, one strategy block each, layout anti-hallucination rules, an image-generation prompt per concept, and a batch CSV.
- Video scenario (+ roast) - narrative spine selection with the why, beat table with timing budgets, persona-locked VO, shot-by-shot storyboard - and an adversarial panel verdict (provenance, persona, feed-realism, copy economy, pacing) applied before you ever see the scenario.
- Competitor spy - live Meta Ad Library research: angle × format coverage maps, white-space cells, longevity-ranked top performers, and one faithful rip-off brief per distinct competitor template.
- Funnel diagnosis - the production 7-step bottleneck walk, ending in a verdict: the problem is the AD, the FUNNEL, the OFFER, or your TARGETS - plus the 15-factor funnel scoring.
- Launch plan - readable test cells (one angle × audience hypothesis per ad set), budgets vs caps, UTMs, and day-0/3/7 checkpoints.
Hardening
- Nine registry tools wired (live Ad Library research, copy validation, angle scoring, CSV generation, and more)
- Structured MODE announcement on every run, verified by an automated eval harness
- Three new one-click presets on the agent page: video scenario, competitor spy, funnel diagnosis