# InjectScan

> InjectScan is a freemium MCP server for ai models and llm tooling. Prompt-injection scanner for text, web pages and MCP tool definitions, paid per call. It runs as a hosted remote server, so there is nothing to install locally.

- **Kind:** MCP server
- **Pricing:** Freemium
- **Platforms:** API, Web
- **Website:** https://injectscan.agentgate.workers.dev
- **Repository:** https://github.com/agentgate-builder/x402-prompt-injection-scan
- **MCP Registry name:** io.github.agentgate-builder/x402-prompt-injection-scan
- **Categories:** AI models and LLM tooling, security
- **Listing updated:** 2026-11-15

## About

Three tools cover what an agent might read before acting. scan_text checks raw text up to 100,000 characters, scan_url fetches a page and also inspects text hidden in comments, CSS-hidden elements and alt, aria or meta attributes, and scan_tools checks up to 100 MCP tool definitions for tool poisoning. Each result has a 0 to 100 risk score, a verdict, categories and the exact spans that matched.

Detection is rule based and every rule, weight and pattern is public at /v1/rules. It covers instruction override in ten languages, jailbreak personas, exfiltration through markdown images, hidden zero-width and tag characters, encoded payloads and fake delimiter tokens.

The vendor is open about limits. It is a pattern scanner, so paraphrased attacks can pass: published benchmark recall is 0.27 on the deepset set and 0.66 on Gandalf attacks, with false positives of 1.3% on Wikipedia pages and 5.6% on GitHub READMEs. Text sent is scanned in memory and not stored. The free tier gives 30 full calls a day per IP, then $0.003 per call in USDC over x402.

## Features

- **Three scan tools:** scan_text, scan_url and scan_tools cover raw text, pages and MCP tool definitions.
- **Hidden text checks:** In URL mode it also inspects comments, CSS-hidden elements and alt, aria and meta text.
- **Public rules:** Every rule, weight and pattern is published at /v1/rules.
- **Explainable results:** Returns a 0-100 score, verdict, categories and the exact matching spans.
- **Pay per call:** 30 free full calls a day per IP, then $0.003 per call in USDC over x402.
- **Stated benchmarks:** Publishes recall and false-positive figures, including weak spots.

## Good for

- Screen a fetched web page before an agent follows instructions in it
- Vet a third-party MCP server's tool descriptions for poisoning
- Check emails or documents for hidden instructions
- Add a safety layer to an agent pipeline next to least-privilege tools

## Install

### Claude Code (Terminal)

```sh
claude mcp add --transport http x402-prompt-injection-scan https://injectscan.agentgate.workers.dev/mcp
```

### Claude Desktop (Settings → Connectors → Add custom connector)

```sh
https://injectscan.agentgate.workers.dev/mcp
```

### Cursor (~/.cursor/mcp.json)

```json
{
  "mcpServers": {
    "x402-prompt-injection-scan": {
      "url": "https://injectscan.agentgate.workers.dev/mcp"
    }
  }
}
```

### VS Code (.vscode/mcp.json)

```json
{
  "servers": {
    "x402-prompt-injection-scan": {
      "type": "http",
      "url": "https://injectscan.agentgate.workers.dev/mcp"
    }
  }
}
```

### Windsurf (~/.codeium/windsurf/mcp_config.json)

```json
{
  "mcpServers": {
    "x402-prompt-injection-scan": {
      "serverUrl": "https://injectscan.agentgate.workers.dev/mcp"
    }
  }
}
```

### Other clients (your client's MCP settings (most use this shape))

```json
{
  "mcpServers": {
    "x402-prompt-injection-scan": {
      "type": "http",
      "url": "https://injectscan.agentgate.workers.dev/mcp"
    }
  }
}
```

## Tools

- `scan_text`: Scans text up to 100,000 characters for prompt injection.
- `scan_url`: Fetches a page and scans its visible and hidden text.
- `scan_tools`: Scans up to 100 MCP tool definitions for poisoning.

## FAQ

### How much does it cost?

Free for 30 full calls a day per IP (10 a minute), then $0.003 per call in USDC over x402. No account or API key.

### How accurate is it?

It is a pattern scanner. Published figures include recall 0.81 on a jailbreak set but 0.27 on deepset/prompt-injections, and 5.6% false positives on GitHub READMEs.

### Is my text stored?

The service says text you send is scanned in memory and not stored.

### How do I connect it?

Add https://injectscan.agentgate.workers.dev/mcp as a remote Streamable HTTP server, for example with claude mcp add --transport http.

### Is InjectScan free?

InjectScan has a free plan, with paid plans for more usage or features.

### How do I add InjectScan to Claude Code?

Run this command in your terminal: claude mcp add --transport http x402-prompt-injection-scan https://injectscan.agentgate.workers.dev/mcp

### What are the best alternatives to InjectScan?

The closest alternatives to InjectScan are REA, Claude Code Ultimate Guide, Agent Roast, Midplane and Leash.

## Alternatives

- [REA](https://ai-powered.com/mcp/rea): REA is an open-source MCP server for developer tools.
- [Claude Code Ultimate Guide](https://ai-powered.com/mcp/claude-code-ultimate-guide): Claude Code Ultimate Guide is an open-source MCP server for ai models and llm tooling.
- [Agent Roast](https://ai-powered.com/mcp/agent-roast): Agent Roast is an open-source MCP server for developer tools.
- [Midplane](https://ai-powered.com/mcp/midplane): Midplane is an open-source MCP server for databases.
- [Leash](https://ai-powered.com/mcp/leash): Leash is a free MCP server for security.
- [Tinplate Edge AI Safety, Security & EU Compliance](https://ai-powered.com/mcp/tinplate-edge-ai-safety-security-and-eu-compliance): Tinplate Edge AI Safety, Security & EU Compliance is a freemium MCP server for security.
- [SecureCloud](https://ai-powered.com/mcp/securecloud): SecureCloud is a free MCP server for security.
- [Eikino](https://ai-powered.com/mcp/eikino): Eikino is a freemium MCP server for security.

---
Source: https://ai-powered.com/mcp/injectscan-2
