Roy Azran

Building mcp-ads
1000$ mrr

[Roy Azran @McpAdss]

@DanielSmidstrup 1000$

[Roy Azran @McpAdss]

@jack Any other languages?

[Roy Azran @McpAdss]

Day 4 of asking @marclou to add airwallex API to trustedMRR so I can sell my 1000$ mrr saas.

[Roy Azran @McpAdss]

@pcshipp My friend let ai so it for you , that's why I created https://t.co/ogqcoRMpqo

[Roy Azran @McpAdss]

Day 3 of asking @marclou to add airwallex API to trustedMRR so I can sell my 1000$ mrr saas.

[Roy Azran @McpAdss]

@danielpopa_me Good post. Most of the "AI got me banned" stories skip the actual suspension reason. Read-only first, writes behind confirmation and staying inside rate limits covers almost all of it.

[Roy Azran @McpAdss]

@pranav_speech Cross-platform is where it really pays off, since each platform grades its own homework. Next step I'd add: have it check the platform conversions against your actual backend orders or leads, not just what Ads Manager reports.

[Roy Azran @McpAdss]

@dannyveigatx Love that it read 45 days of context before touching anything. Most agent failures I see come from acting on last week's numbers with zero history. Curious what it does the first week it misses the 5/week target: propose spend changes or just flag?

[Roy Azran @McpAdss]

@apranevicius Exactly the bottleneck. One of our users paid a marketer ₪2,000/mo for one new creative a day; now his AI agent does it automatically via https://t.co/ogqcoRMpqo for $49/mo. A strategist still picks the angles, but volume stops being the constraint.

[Roy Azran @McpAdss]

@stacy_siz Saving this. One thing I wish more beginners checked in week one: the placement breakdown. Audience Network can quietly eat a small budget before you notice. Looking forward to part 2.

[Roy Azran @McpAdss]

@aalanazarr Not inherently, but daily uploads into one CBO spread learning thin and the old winner usually keeps the spend anyway. Batching new ads weekly, or a separate testing campaign, gives each creative a fair read before you judge it.

[Roy Azran @McpAdss]

@Serhii_Klym @stacy_siz No need for a new campaign. Edit placements at the ad set level to manual, FB + IG only. Audience Network often soaks up small budgets with cheap low-intent traffic, especially for apps. It'll re-enter learning, but that beats paying for junk.

[Roy Azran @McpAdss]

@AnmolKulsh57726 @RobbyFrank Rule of thumb: optimize for trials from day one only if you'll get ~10+ a week; otherwise start on installs and switch once volume is there. $20-30/day is fine to learn, just give it ~2 weeks before judging. Tier 1 only if your pricing needs it.

[Roy Azran @McpAdss]

@brinleyc2 For SaaS I'd start narrow: bottom-of-funnel keywords (your category + "software", "[competitor] alternative"), phrase/exact match, one landing page per theme, real signups tracked. Check the search terms report every few days and add negatives.

[Roy Azran @McpAdss]

@ajith_io Boosting proven organic posts is a smart first test. With $500: one campaign, Sales or Leads objective (not Engagement), 3-5 posts in one ad set, broad audience. Don't judge before ~3 days, and cut posts that spend 2x your target CPA with nothing.

[Roy Azran @McpAdss]

@r1VeN2k Yes, but scoped. Pausing and drafting can run freely since they only save money or wait for review. Anything that raises spend or goes live should wait for a human click. Your point about warnings not routed to a person is the one most setups miss.

[Roy Azran @McpAdss]

@ecommcowboy @PhilKiel Agree with Phil on SOPs. What I'd automate into BFCM: morning spend/CPA checks vs target, flagging ads past their fatigue point, drafting new variations. Launches and budget jumps still get a human yes. Building https://t.co/ogqcoRMpqo around that.

[Roy Azran @McpAdss]

Agree with Phil on SOPs. What I'd automate into BFCM: morning spend/CPA checks vs target, flagging ads past their fatigue point, drafting new variations. Launches and budget jumps still get a human yes. Building https://t.co/ogqcoRMpqo around that.

[Roy Azran @McpAdss]

@Hendrixoncode Nice setup. Next one I'd hand over: creative refresh. Have it flag ads with falling CTR or rising frequency and draft 3 variations from your winners for you to approve. That's the piece I built https://t.co/ogqcoRMpqo around, for Google + Meta.

[Roy Azran @McpAdss]

@ServiceAdsGuy From what I've seen it leans on Meta's general guidance, not a real read of your account. If you want answers grounded in your own numbers, connect the account to Claude/ChatGPT via an MCP (Meta has one, mine is https://t.co/ogqcoRMpqo) and have it cite the rows.