Managing Meta Ads campaigns often means spending 20 minutes or more daily clicking through dashboards, checking performance trends, tweaking budgets, and refreshing creatives. meta-ads-kit offers a different route: a Shell-based AI agent system that automates this entire loop. It monitors spend pacing, detects creative fatigue, generates new ad copy matched to existing images, and directly updates campaigns via Meta’s Graph API — all without opening the Ads Manager UI.
what meta-ads-kit does and how it works
meta-ads-kit is an open-source AI ad manager built around an automation loop that replaces much of the manual work advertisers do daily. It is implemented primarily as a Shell CLI tool orchestrated by an AI agent framework called OpenClaw, combined with social-cli for social media API interactions.
The core of the system revolves around answering five daily questions that summarize campaign health:
- Are we pacing spend correctly?
- Which campaigns are active?
- What are the 7-day performance trends?
- Which ads or campaigns are winners or losers?
- Are there signs of creative fatigue?
These questions guide the AI agent’s workflow, which fetches performance data, analyzes it for fatigue signals, decides on budget shifts, and generates new ad copy that is matched to the image creatives already in use. The AI-generated copy is then pushed directly to Meta Ads via the Graph API.
This closed-loop automation means marketers no longer have to manually open Ads Manager, check multiple metrics, and write new copy themselves. Instead, the system outputs a concise 2-minute briefing and automates updates.
Under the hood, meta-ads-kit is a Shell-based CLI tool that leverages:
- OpenClaw: An AI agent framework that orchestrates the workflow logic, manages conversational prompts, and maintains state.
- social-cli: A Node.js CLI tool that handles API calls to social platforms, here used to interact with Meta’s Graph API.
- Configuration via JSON files and prompt templates to tune the AI’s decision-making.
The repo’s design favors modularity and transparency, letting users adjust benchmarks and prompts to fit their campaigns.
the AI-driven closed loop: automation with tradeoffs
What stands out technically is the closed-loop system that combines monitoring, analysis, creative generation, and deployment driven by AI. This is not just a reporting tool; it actively manages campaign adjustments based on AI’s interpretation of data.
The tradeoff is clear: the system depends heavily on the quality and stability of Meta’s Graph API and on the AI’s ability to interpret ambiguous signals like fatigue — a subjective concept.
Implementing this as a Shell CLI tool orchestrated by OpenClaw and social-cli means the codebase is surprisingly lean and accessible but sacrifices some UI polish and real-time interactivity.
The code quality reflects a pragmatic approach: shell scripts and JSON configurations handle data flow, while AI prompts encapsulate the decision logic. This makes it easy to tweak but requires comfort with AI prompt engineering and CLI tools.
Another strength is the use of the “5 Daily Questions” framework, a clever abstraction that distills complex campaign health into manageable insights. This framework efficiently guides the AI agent’s tasks and reduces cognitive load for users.
That said, the system assumes a level of familiarity with AI agents, CLI environments, and Meta Ads APIs. It’s not a plug-and-play solution for all marketers but rather a tool for technically inclined users who want to automate tedious workflows.
quick start
git clone https://github.com/themattberman/meta-ads-kit.git
cd meta-ads-kit
# Install social-cli (the engine under the hood)
npm install -g @vishalgojha/social-cli
# Install OpenClaw
npm install -g openclaw
These commands set up the core dependencies. After installation, users configure the system via JSON benchmarks and conversation prompt files included in the repo. Running the CLI initiates the AI agent’s daily workflow.
verdict: who should use meta-ads-kit?
meta-ads-kit is a niche but practical tool for developers and marketers comfortable with AI tooling and command-line environments who want to automate Meta Ads management. It replaces a tedious manual process with an AI-driven closed loop that intelligently monitors campaigns, detects creative fatigue, and generates copy tied to existing creatives.
Its limitations include reliance on Meta’s API stability, the challenge of tuning AI prompts for accurate fatigue detection, and the absence of a polished UI. It’s not for casual users or those expecting a turnkey SaaS dashboard.
But if you want to build or customize an automated Meta Ads manager that integrates AI agent orchestration with direct API control, this repo is a solid foundation. The approach of distilling campaign health into five daily questions is a useful pattern for anyone automating ad workflows.
The code is surprisingly straightforward given the complexity of the problem space, and the CLI-based AI agent orchestration is a clean, modular design that’s worth understanding even if you don’t adopt it wholesale.
→ GitHub Repo: TheMattBerman/meta-ads-kit ⭐ 221 · Shell