← Files AdAgntARCHIVED FILE

plugin.json

7.38 KB · Nov 20, 2026 · 14:45 UTC

↓ Download file

See the change to this file →

{
  "$schema": "https://agent-plugins.org/schemas/1.0.0/plugin.schema.json",
  "name": "app-6a86b53a9668819188845b69c4e4aacc",
  "version": "1.0.2",
  "description": "Run ads from the AI you already use: 352 tools across Google, Meta, TikTok, Amazon, AppLovin, Reddit, X and Snapchat Ads. Nothing goes live without your approval.",
  "author": {
    "name": "AdAgnt",
    "email": "support@adagnt.com",
    "url": "https://adagnt.com"
  },
  "homepage": "https://adagnt.com",
  "license": "Proprietary",
  "keywords": [
    "ads",
    "google ads",
    "meta ads",
    "tiktok ads",
    "amazon ads",
    "reddit ads",
    "x ads",
    "snapchat ads",
    "paid media",
    "marketing"
  ],
  "extensions": {
    "com.openai": {
      "interface": {
        "displayName": "AdAgnt",
        "shortDescription": "Run ads in plain English",
        "longDescription": "AdAgnt is the ads manager you talk to. Say what you want in plain English \u2014 \"find the spend that isn't converting\", \"build a $20/day search campaign for my app\", \"compare Google and Meta this month\" \u2014 and AdAgnt does the work across Google Ads, Meta Ads, TikTok Ads, Amazon Ads, AppLovin, Reddit Ads, X Ads and Snapchat Ads. 352 tools, one conversation.\n\nWhat you can do\n\u2022 Connect a platform right from the chat. AdAgnt shows every platform with a Connect button that opens that platform's own sign-in. AdAgnt never sees your password.\n\u2022 Research keywords with real search volumes and bid ranges, then build a complete campaign \u2014 budget, targeting, ads and tracking \u2014 ready to launch.\n\u2022 Compare performance across accounts and platforms in one answer, find wasted spend, and rebalance budgets toward what converts.\n\u2022 Run A/B tests judged on a real significance test, check your numbers against industry benchmarks, and get a scale / keep / fix / kill verdict per campaign.\n\u2022 Let monitors check your numbers every day, and let an autopilot sweep file suggested changes into an approval queue that waits for you.\n\nYou stay in control. On a connected account every change comes to you as a proposal on a card in the chat: nothing changes until you approve it, and AdAgnt reads the platform back afterwards to confirm what changed. Every change is recorded with why it was made.\n\nStart in sandbox. Every account begins in a deterministic simulation with realistic campaigns and 90 days of metrics, so you can try everything with zero risk and no ad account. Nothing there touches a live platform and no money can move.\n\nLive connections today: Google Ads, Amazon Ads, Reddit Ads and X Ads; AppLovin connects live with its Report Key. Meta, TikTok and Snapchat work in the sandbox while their API applications are in review. The app shows each platform's status before you connect anything.\n\nRevenue-based verdicts are modelled from your conversions and average order value and labelled as estimates. AdAgnt does not ingest orders from GA4, Shopify, Stripe or Klaviyo.",
        "developerName": "AdAgnt",
        "category": "Business & Operations",
        "capabilities": [
          "Read",
          "Write"
        ],
        "websiteURL": "https://adagnt.com",
        "supportURL": "https://mcp.adagnt.com/support",
        "privacyPolicyURL": "https://mcp.adagnt.com/privacy",
        "termsOfServiceURL": "https://mcp.adagnt.com/terms",
        "defaultPrompt": [
          "Show my ad connections",
          "How did my campaigns perform in the last 30 days?",
          "Find wasted spend and propose fixes for me to approve"
        ],
        "brandColor": "#1F7E79",
        "composerIcon": "./assets/wordmark-512.png",
        "logo": "./assets/wordmark-512.png"
      },
      "review": {
        "test_cases": {
          "positive": [
            {
              "description": "Connections card \u2014 see what is linked and connect a platform from the chat",
              "prompt": "Which ad accounts are connected? I want to connect Google Ads.",
              "tools_triggered": "get_connections_status",
              "expected_behavior": "An inline AdAgnt card lists all eight platforms with a status pill (Live or Sandbox) and a button. Connect opens app.adagnt.com/dashboard/accounts?connect=google in a new tab where Google's own sign-in starts; Check again re-runs the tool. Read-only."
            },
            {
              "description": "Campaign creation \u2014 write path",
              "prompt": "Create a Google search campaign called 'Mobile Bike Repair \u2014 Portland' with a $75/day budget for https://www.example.com/bike-repair targeting Portland, OR. One ad group with mobile bike repair keywords.",
              "tools_triggered": "create_search_campaign",
              "expected_behavior": "The assistant presents the plan (headlines, descriptions, keywords, budget, targeting). On the sandbox the campaign is created PAUSED in the simulator and the result gives its id and structure; it then appears in list_campaigns. On a connected Google Ads account the same call returns an approval card and nothing is created until the account owner approves it. No money moves in the sandbox."
            },
            {
              "description": "Cross-platform performance analysis",
              "prompt": "With AdAgnt, show me how all my ad campaigns performed in the last 30 days and tell me which platform had the best return.",
              "tools_triggered": "get_campaign_performance, get_meta_campaign_performance, get_tiktok_campaign_performance, amazon_get_sp_campaigns_report",
              "expected_behavior": "Per-platform spend, clicks, conversions and ROAS, then a comparative conclusion naming the best-returning platform. Read-only."
            },
            {
              "description": "Wasted-spend audit \u2014 read-only",
              "prompt": "Use AdAgnt to find where I'm wasting ad spend and tell me what to change \u2014 don't change anything yet.",
              "tools_triggered": "analyze_wasted_spend",
              "expected_behavior": "Zero-conversion spenders and inefficiencies with recommendations. No write tool is invoked."
            },
            {
              "description": "Intelligence layer \u2014 revenue-based verdicts, labelled as estimates",
              "prompt": "In AdAgnt, set my revenue assumptions to Shopify with a $90 average order value, then show me ROAS estimates per campaign and which ones to scale or kill.",
              "tools_triggered": "set_revenue_assumptions, get_roas_estimates",
              "expected_behavior": "Per-campaign ROAS with scale / keep / fix / kill verdicts, explicitly presented as estimates modelled from conversions and average order value (the revenue_basis field says so). Calling them verified revenue would be a failure."
            }
          ],
          "negative": [
            {
              "description": "General marketing knowledge \u2014 no account action",
              "prompt": "What's the difference between CPC and CPM bidding?"
            },
            {
              "description": "Unrelated domain",
              "prompt": "Plan me a 3-day itinerary for Rome."
            },
            {
              "description": "Adjacent but out of scope (organic, not paid)",
              "prompt": "Write an SEO-optimized blog post about bike chain maintenance."
            }
          ]
        },
        "demo_recording_url": "https://adagnt.com/video/adagnt.mp4",
        "commerce": false
      },
      "apps": "./.app.json"
    }
  }
}

SHA-256: 1adf4439f76fa096ebcdc65cb4225d1685ac92afd0a4fc62120269cc7d848d6c