Model Context Protocol (MCP) server that lets AI models query and manage Meta (Facebook/Instagram) advertising accounts via a uniform tool interface.
https://github.com/pipeboard-co/meta-ads-mcpYou know the drill: campaign performance drops, you tab over to Facebook Ads Manager, dig through metrics, try to figure out what's happening, then make educated guesses about fixes. Meanwhile, your AI assistant sits there useless because it can't see your ad data.
Meta Ads MCP changes that. Now your AI can directly query your Facebook and Instagram ad accounts, analyze performance, spot issues, and suggest optimizations - all without you leaving your current workflow.
Every developer running ads hits the same friction: ad platforms live in isolated silos. Your AI tools can help with code, writing, and analysis, but the moment you need to understand why your app's acquisition campaigns aren't performing, you're back to manual dashboard clicking.
You end up with:
This MCP server connects any AI model to Meta's advertising API through a clean, standardized interface. Instead of learning Facebook's marketing API or building custom integrations, you get 22 ready-to-use tools that handle everything from campaign analysis to creative optimization.
Real Example: Instead of opening Ads Manager to check why your iOS app campaign CTR dropped 30%, you ask Claude: "Analyze my campaign performance for the last week and tell me what's causing the CTR drop." It pulls the data, identifies the issue (maybe iOS 14.5 attribution changes), and suggests specific targeting adjustments.
Campaign Optimization: "Look at my ad spend across all campaigns this month. Which ones are underperforming and how should I reallocate budget?"
Creative Analysis: Your AI downloads ad images, analyzes creative performance, and suggests improvements: "Your video ads are getting 40% higher engagement than static images, but you're spending 70% of budget on static creatives."
Automated Monitoring: Set up AI-powered alerts for performance changes. When spend efficiency drops or CPA spikes, get actionable recommendations immediately instead of discovering issues days later.
A/B Test Analysis: "Compare the last two weeks of Campaign A vs Campaign B performance. Break down by demographics and tell me which audience segments are driving the difference."
Budget Scheduling: Create dynamic budget adjustments for high-demand periods without manual intervention.
Remote MCP (Recommended for most developers): No setup required. Add one line to your Claude or Cursor config, authenticate with Facebook, and start analyzing campaigns immediately. Perfect when you want the functionality without infrastructure overhead.
Self-Hosted: Full control over the MCP server. Clone the repo, set your Meta API tokens, run locally. Better for teams needing custom configurations or handling sensitive campaign data internally.
Both options use the same 22 MCP tools - from basic account queries to advanced creative management and insights analysis.
For Claude Users: Add the remote MCP URL in your Claude settings, connect your Facebook account, and ask: "Show me my top 5 campaigns by spend this month and their key metrics."
For Self-Hosting:
uvx meta-ads-mcp
export PIPEBOARD_API_TOKEN=your_token
# Add to your MCP client config
For Cursor: Add the remote MCP configuration to your mcp.json and you're analyzing campaigns in your editor.
The fastest way is Remote MCP - you'll be getting AI insights on your campaigns within 2 minutes of setup.
If you're running ads for your apps, SaaS products, or consulting clients, this eliminates the most tedious part of campaign management. Instead of spending time gathering and interpreting data manually, you focus on strategic decisions while AI handles the analysis.
You're already using AI for code review, documentation, and problem-solving. Adding advertising intelligence to that workflow is the logical next step - especially when the setup takes under 5 minutes.