AI media buying copilot

Catch the anomaly before it eats your margin.

Altaviz watches every campaign across Meta, Google, Taboola, and TikTok — finds what broke, prices it in dollars, and drafts the fix for your approval.

5
anomaly detectors
7
agent + MCP tools
$6,340
per day surfaced in the demo
1Open the demo

The engine has already found every problem in the account — priced per day.

2Expand a finding

Statistical evidence, no LLM guesses: the Google outage, the fatigued solar ad.

3Ask, then approve

“What should I kill today?” — approve the copilot’s actions in the queue.

Media buying teams lose money in the gap between broken and noticed.

A fatigued creative quietly doubles CPA. A tracking outage burns a day of spend while dashboards look normal. A winner sits budget-capped because scaling it was nobody's job that week.

At affiliate scale that detection lag is a permanent tax on ROI. Dashboards show numbers — not what changed, what it costs, and what to do. Altaviz closes that gap.

What it does

Detect

Statistical detection over every campaign — creative fatigue, CPA drift, spend spikes, tracking outages, underfunded winners. Every finding priced in $/day.

z-scores · trend slopes · significance gates

Decide

A Claude copilot grounded in the same tools as the dashboard. Ask "what should I kill today?" — get dollars, evidence, and the why.

claude · tool-calling agent

Act, approved

Typed actions with exact platform-API params, queued for human approval. An agent never spends money unattended.

human-in-the-loop · never auto-executes

The same tools, from Claude or Cursor.

Everything behind the copilot is an MCP server at /api/mcp. Plug the account into the AI tools your team already uses.

Claude Code: claude mcp add --transport http altaviz https://altaviz.vercel.app/api/mcp

{
  "mcpServers": {
    "altaviz": {
      "command": "npx",
      "args": ["-y", "mcp-remote",
        "https://altaviz.vercel.app/api/mcp"]
    }
  }
}