# Optifeed Radar MCP server

Canonical HTML documentation: https://radar.optifeed.com/mcp/

The Optifeed Radar MCP server exposes brand and product AI visibility capabilities to MCP clients over local stdio. It runs on the user's machine, uses the user's provider keys, and saves results locally.

## Connect the published package

Run the published `optifeed-radar@0.1.0` package:

```json
{
  "mcpServers": {
    "optifeed-radar": {
      "command": "npx",
      "args": ["-y", "--package=optifeed-radar", "optifeed-mcp"]
    }
  }
}
```

Contributors can instead clone and build the repository:

```bash
git clone https://github.com/optifeed/optifeed-radar.git
cd optifeed-radar
npm install
npm run build
```

Then configure the client to run:

```text
node /absolute/path/to/optifeed-radar/dist/mcp/index.js
```

## Tools

### `audit_store`

Run a free readiness audit for a domain. Checks AI crawler access, llms.txt, structured data, metadata, and sitemap. No AI calls.

Required argument: `domain`.

### `check_visibility`

Ask configured engines buyer questions and estimate whether the brand is recommended.

Arguments:

- `domain` — required brand site.
- `engines` — optional array of `openai`, `anthropic`, `gemini`, `perplexity`.
- `quick` — optional smaller prompt pack.
- `max_cost` — optional USD spend cap; default $0.50.

### `generate_buyer_queries`

Discover the brand and generate its intent-matched buyer-question pack. Required argument: `domain`. This normally makes a small number of model calls.

### `shopping_check`

Check products explicitly supplied by the user, best first. The order is treated as the merchant's ranking and compared with observed AI recommendation order. There is no catalog discovery.

Arguments:

- `domain` — required store site.
- `products` — required list of up to ten names or objects with `name`, optional `aliases`, and optional `descriptor`.
- `engines` — optional engine list.
- `max_cost` — optional USD spend cap; default scales at $0.20 per product.

### `get_snapshot_diff`

Compare the two latest locally saved brand runs. Required argument: `domain`. No AI calls.

## Example prompts

- “Run a free AI-readiness audit on yourbrand.com.”
- “Show me the buyer questions yourbrand.com should be visible for.”
- “Check yourbrand.com's AI visibility in quick mode and cap it at $0.20.”
- “Check yourbrand.com using OpenAI and Perplexity only.”
- “Check these products, best first: Product A, Product B.”
- “What changed since the last visibility run?”

## Safety

Begin with `audit_store` because it requires no keys and spends nothing. Paid tools use provider credit from locally configured keys. Use an explicit `max_cost` for an unattended agent run. Scores are estimates from sampled answers and vary between runs.

Version 0.1.0 runs locally over stdio. A hosted remote MCP endpoint is roadmap work, not a current capability.

Repository: https://github.com/optifeed/optifeed-radar