About the Author

I'm Aleksej Kruminsh, a performance marketing operator and founder of MarketingRoutine. The byline on every article in this guide links back here because the disclosed-self-listing model only works when readers can audit who is making which call. This page is the audit surface.

What I do

Day to day, I run Meta Ads and paid search across direct-brand and agency contexts. That work spans multiple markets, each with its own consumer norms, creative idioms, and pricing sensitivity. I write ad copy, set targeting, build creatives, and increasingly run those decisions through MCP-based AI workflows in Claude Desktop and ChatGPT.

The tools I rely on weekly: Meta Ads Manager, the Meta Marketing API directly via curl when the UI is unhelpful, Meta's own ads MCP server (since the April 2026 connectors launch), and MarketingRoutine for Meta-only workspace and creative performance workflows. I've also installed Pipeboard locally for code-audit work — when a question is "what is actually being sent to Meta?", a self-host server I can read line-by-line is the correct answer.

Real operator context

Running paid performance in-house taught me how to evaluate a tool not by what its marketing claims but by what survives sustained campaign operations across markets, budgets, and creative cycles. That perspective is what I try to carry into every comparison page in this guide. When I write that a particular MCP server "works" or "doesn't," that judgment is grounded in real campaign operations, not a sandbox demo.

Independent agency context

I also run paid-media operations through Marketing Hackers, a Latvia-based growth agency. I don't name individual clients for confidentiality reasons, but the agency context shapes how I evaluate things: per-client token isolation, workspace separation, observability into account changes, and rate-limit governance across multiple ad accounts matter in a way they don't for a single-brand setup. The "best for agencies" use-case page in this guide is grounded in those concerns.

Why this guide exists

Most "Meta MCP" content I encountered when the third-party MCP category took off was either thin vendor marketing or fragmented GitHub READMEs missing the operational context. Neither answers the actual questions a working marketer asks at 11pm before installing a new tool: Which one do I install? Will it ban my ad account? How does it compare to Meta's own server? What can I trust to run unattended overnight?

This guide tries to answer those questions honestly, with hands-on testing on real campaigns, primary-source citations, and a public-record changelog when I get something wrong.

The MarketingRoutine disclosure

Direct disclosure, repeated on every comparison page in case anyone arrives via search rather than from this About: I work with the team that publishes MarketingRoutine. It appears in the comparison cluster of this guide alongside Pipeboard, Madgicx, Ryze AI, brijr/meta-mcp, and Meta's own server. The same 13-criterion rubric applies to MarketingRoutine and to every other product evaluated. The comparison pages name the criteria where MarketingRoutine loses — SaaS-only (loses to Pipeboard self-host), Meta-only (loses to Madgicx multi-platform), younger product than several established commercial alternatives — without softening the language.

If reading this guide makes you choose a different MCP server than MarketingRoutine for your use case, the guide is doing its job.

My editorial commitments

• Primary sources for every factual claim. GitHub repos, official Meta and MCP-spec documentation, vendor websites — never blog posts or AI-generated content as primary sources. Citations link directly to the source so you can verify without leaving the page.
• Disclosed ratings against my own product. MarketingRoutine is rated on the same rubric as every competitor. The comparison pages name where it loses, in plain language, without burying the criticism.
• Update what's wrong. When a comparison rating becomes inaccurate — because Meta updated their MCP, a competitor shipped new features, or my testing missed something — I update the article, bump lastVerified, and log the change in the changelog.
• No paid placement, no affiliate links. No vendor pays to be included, excluded, or rated higher. The 13-criterion rubric is the same regardless of who I work with.

Connect and corrections

The fastest way to reach me about Meta ads operations, MCP-based automation, or paid-media work at scale is LinkedIn.

Found something wrong in an article? Email hello@marketingroutine.ai with the specific paragraph and the correct fact. Corrections land on production with a lastVerified bump and a changelog entry naming what changed and when.