Markifact MCP server
{
"mcpServers": {
"markifact": {
"type": "http",
"url": "https://api.markifact.com/mcp"
}
}
}{
"mcpServers": {
"markifact": {
"type": "http",
"url": "https://api.markifact.com/mcp"
}
}
}{
"mcpServers": {
"markifact": {
"type": "http",
"url": "https://api.markifact.com/mcp"
}
}
}{
"servers": {
"markifact": {
"type": "http",
"url": "https://api.markifact.com/mcp"
}
}
}{
"mcpServers": {
"markifact": {
"type": "http",
"url": "https://api.markifact.com/mcp"
}
}
}file .mcp.json (in your project root)file claude_desktop_config.jsonfile ~/.cursor/mcp.jsonfile .vscode/mcp.jsonfile ~/.codeium/windsurf/mcp_config.jsonversion 1.0.0
What it does
Markifact exposes over 1000 operations across advertising platforms (Google Ads, Meta, TikTok, LinkedIn, Microsoft, Reddit, Pinterest, Snapchat, Amazon, DV360), analytics (GA4, BigQuery, Google Search Console), e-commerce (Shopify), CRM (HubSpot, Klaviyo), and messaging (Slack, WhatsApp). It is for marketing teams who want AI clients like Claude or ChatGPT to manage campaigns, analyze performance, and execute marketing tasks with human approval required for all writes.
- Create and launch ad campaigns across Google Ads, Meta, TikTok, or LinkedIn with paused status for review before going live.
- Audit ad account spending to identify wasted budget and apply negative keywords or pause underperforming ads.
- Query analytics data from GA4 or Google Search Console and rotate fatigued creative across active campaigns.
Tools this server exposes
| Tool | Does |
|---|---|
| launch-google-search | Create a paused Google Search campaign with ad groups, keywords, and ads. |
| launch-pmax | Create a paused Performance Max campaign with asset group. |
| launch-meta-campaign | Build a paused Meta campaign, ad set, and ads end to end. |
| edit-meta-creative | Update URL, copy, or CTA on a live Meta ad. |
| rotate-creative | Pause fatigued ads and create paused variants for launch. |
| negative-keyword-sweep | Find search terms wasting budget and block them. |
Before you enable it
An MCP server runs with whatever access you give it. Start with a development account rather than production, and check which tools the server exposes before you let an agent call them unattended.