After months of manually managing 37+ client ad accounts on Meta using curl and scripts, a marketing expert has developed and open-sourced an innovative Meta Ads MCP (Meta Conversion Platform) server. This server allows users to manage all aspects of Meta Ads directly from Claude Code, significantly boosting operational efficiency.
Named “KonQuest Meta Ads MCP,” this open-source project (MIT licensed) provides 57 core tools specifically designed for Meta Ads management via Claude Code, covering key functionalities:
- Campaign Management: Create, read, update campaigns, ad sets, ads, and creatives. Users can duplicate entire campaigns with all their ads in one command.
- Multi-Asset Ads: Create a single ad with both video (9:16 + 1:1) and static images (1:1 + 4:5 + 9:16) using
asset_feed_spec, eliminating the need to create separate ads per format. - Targeting Toolkit: Search interests, behaviors, demographics, and geo-locations. Estimate audience sizes before creating ad sets.
- Pixel Diagnostics: Check pixel health, verify event firing, test event sends, and diagnose tracking issues on any URL.
- Catalog and DPA: Manage product catalogs, feeds, and product sets for dynamic product ads, with the ability to create and update product sets with filters.
- Setup Checker: Offers 42+ automated checks per account, including token health, pixel status, page connections, Instagram identity, and catalog links.
How it Works in Practice
Managing Meta Ads from Claude Code is designed to be highly efficient and intelligent:
- When instructed, for example, to “Create a conversion campaign for my skincare store targeting women 25-45 interested in beauty,” Claude reads your local “vault” files (brand voice, ICP profiles, angles), validates everything, enforces naming conventions, checks pixel firing, and then automatically creates:
- A PAUSED campaign
- An ad set with targeting
- Ads with creatives
- If asked, “What's the ROAS on campaign C5 this week?” Claude pulls insights, breaks down data by ad set, flags underperformers, and suggests budget shifts.
- For a command like, “Duplicate campaign C3 but change targeting to men 30-50,” Claude duplicates the full structure (campaign + ad sets + ads), applies the new targeting, and maintains consistent naming conventions.
Safety First
The server prioritizes safety, with all write operations supervised:
- Ads Created PAUSED: All created ads are set to PAUSED by default, requiring explicit approval before going live.
- Naming Enforcement: Every object must adhere to predefined naming conventions before the API call is made.
- Validation Pipeline: Compliance, creative specifications, tracking, structure, and operational checks run before every write operation.
- Rate Limiting: Respects Meta API limits with automatic backoff mechanisms.
- Rollback References: Every change is logged, allowing users to undo operations.
- Safety Tiers: Provides sandbox, standard, and production access levels per account.
The Vault System
The server reads local Markdown files, referred to as the “vault,” to access client intelligence. An example structure might be:
your-vault/
01_CLIENTS/my-client/
00-profile.md # Account IDs, pixel, page
02-icp-personas.md # Target audience profiles
04-brand-voice.md # Tone, language, style
05-messaging-house.md # Angles, value props
matrix.md
This “vault” system ensures that the AI agent executes tasks fully aligned with the client's brand voice, target audience characteristics, and marketing strategies.