# adtest-mcp [Health: Active]

**Category:** 🎯 Marketing  
**Repository:** https://github.com/menaker/adtest-mcp  
**GitHub Stars:** 0  
**npm Downloads (last month):** 214  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/adtest-mcp

## Description
AI-only ad scoring for Claude: score image, video or text ads on 13 dimensions before you spend.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "adtest-mcp": {
    "command": "npx",
    "args": ["-y","adtest-mcp"],
    "env": {
      "ADTEST_API_KEY": "",
      "ADTEST_API_BASE": "",
      "ADTEST_POLL_TIMEOUT_MS": ""
    }
  }
}
```

**Requires environment variables:** `ADTEST_API_KEY`, `ADTEST_API_BASE`, `ADTEST_POLL_TIMEOUT_MS` — the values above are empty placeholders; fill in real credentials before running (see the repository for what each one is for).

## Documentation

## What the adtest-mcp MCP server does

The adtest-mcp MCP server exposes AdTest.AI’s AI-only advert analysis through one MCP tool: `analyze_advert`. It evaluates an image, video, or text advertisement across 13 dimensions and returns a detailed scoring report. The response also includes the amount charged and the remaining account balance.

The tool accepts exactly one primary creative input: a public `url`, a local `file_path`, or inline `text`. An optional `brand_url` points to the advertiser’s website and can improve brand or logo detection. Public image and video URLs are downloaded by the service, while local media is uploaded from the machine running the MCP process.

Human-panel validation is not available through this server. That option is provided only in the AdTest.AI web application.

## How it works

Each request creates an asynchronous analysis job through the AdTest.AI developer API. The server receives a job identifier, then checks the job status periodically until it completes or fails. This design supports video processing, which may take several minutes and could exceed the roughly 100-second limit of a single synchronous request.

A successful job is billed once when processing finishes. Failed jobs are not charged. The server converts API failures into readable error text, including invalid credentials, insufficient funds, invalid URLs, unsupported media, rate limits, and failed analysis states.

The default API base is `https://app.adtest.ai/adtest-api`. Requests authenticate with the `x-api-key` header, populated from the configured API key. The default polling limit is 15 minutes, with a setting available for longer videos.

## Setup and configuration

The adtest-mcp MCP server requires Node.js 18 or newer and an AdTest.AI API key. Generate the key from the Developer area of `app.adtest.ai`, and ensure the associated wallet has funds before submitting analysis jobs.

For a package-based setup, configure an MCP client to run `npx -y adtest-mcp` and pass the key as `ADTEST_API_KEY`. A local checkout can instead run `node` against the repository’s `mcp-server/index.js` after installing dependencies with `npm install`.

Supported environment variables are:

- `ADTEST_API_KEY`: API key used for authentication.
- `ADTEST_API_BASE`: Optional API endpoint override, such as a staging endpoint.
- `ADTEST_POLL_TIMEOUT_MS`: Optional polling ceiling in milliseconds.

Restart the MCP client after adding the server configuration. Example prompts can reference a public ad URL, a local video path, or ad copy directly.

## Tools and capabilities

`analyze_advert` supports:

- Image analysis from a public URL or local file.
- Video analysis from a public URL or local file.
- Text-only ad copy analysis.
- Optional advertiser website context through `brand_url`.
- Asynchronous polling for long-running jobs.
- Reporting of analysis results, charges, and remaining balance.

Per-key limits are 30 submissions per minute and 2,000 submissions per day. The service also documents a 60-per-minute limit per IP; polling is handled by the server and has a higher allowance.

## Limitations and notes

The adtest-mcp MCP server performs AI-only analysis; it does not expose the web app’s human-panel validation. A successful analysis costs $1, so the configured wallet must have sufficient funds. Video jobs can take several minutes and may require increasing `ADTEST_POLL_TIMEOUT_MS` beyond the default.

The server depends on reachable public URLs for URL-based media and on supported image or video formats. Requests can fail with `no_creative`, `invalid_url`, `unsupported_media_type`, `too_many_requests`, or `analysis_failed`. The documented `analysis_failed` condition is not billed. Rate limits also apply, so automated workloads should account for the submission quotas.

_Full upstream README: https://allmcps.com/mcp/adtest-mcp/readme_