# VeriSight MCP

> VeriSight MCP is an open-source MCP server for security. Image forensics server that flags AI inpainting, recapture and duplicate claim photos. It runs locally on your machine and is started by your AI client.

- **Kind:** MCP server
- **Pricing:** Open source
- **Platforms:** CLI, self-hosted, API
- **Repository:** https://github.com/adityarajtiwari880/verisight
- **MCP Registry name:** io.github.Adityarajtiwari880/verisight
- **Categories:** [security](https://ai-powered.com/categories/security)
- **Listing updated:** 2026-11-18

## About

VeriSight aims at refund and insurance fraud where a real photo is edited with generative inpainting to fake damage. Its README argues for deterministic checks built on OpenCV instead of neural network weights, and quotes sub-3 ms inspection on CPU. Those speed and accuracy figures are the author's own.

The pipeline combines error level analysis, sensor-noise (PRNU) checks, a 2D Fourier probe for diffusion-grid patterns, JPEG ghost sweeping, perceptual hashing for reused photos, and detection of photos of screens or printouts. A final gate maps the risk score to approve, human review or reject, and the README says evidence dossiers can be written to Amazon S3 and DynamoDB.

It installs from PyPI and can run as an stdio MCP server with uvx, as a Python library, or from a command-line scan command. The README lists seven named forensic tools for MCP clients.

## Features

- **Multi-check pipeline:** Runs ELA, PRNU, Fourier, JPEG ghost and hash checks together.
- **Recapture detection:** Flags photos of screens or printed cards.
- **Duplicate photo detection:** Perceptual hashing finds reused claim images.
- **Risk-based verdicts:** Maps risk to approve, review or reject.
- **No model weights:** Relies on OpenCV rather than neural networks.
- **CLI and Python SDK:** Scan images from a terminal or code.

## Good for

- Screen refund-claim photos for AI-edited damage
- Spot reused photos across insurance claims
- Check coupon or invoice screenshots for tampering
- Add a forensic step to an agent that triages claims

## Install

### Claude Code (Terminal)

```sh
claude mcp add verisight -- uvx verisight
```

### Claude Desktop (claude_desktop_config.json)

```json
{
  "mcpServers": {
    "verisight": {
      "command": "uvx",
      "args": [
        "verisight"
      ]
    }
  }
}
```

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

```json
{
  "mcpServers": {
    "verisight": {
      "command": "uvx",
      "args": [
        "verisight"
      ]
    }
  }
}
```

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

```json
{
  "servers": {
    "verisight": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "verisight"
      ]
    }
  }
}
```

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

```json
{
  "mcpServers": {
    "verisight": {
      "command": "uvx",
      "args": [
        "verisight"
      ]
    }
  }
}
```

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

```json
{
  "mcpServers": {
    "verisight": {
      "command": "uvx",
      "args": [
        "verisight"
      ]
    }
  }
}
```

## Tools

- `verisight_inspect_claim`: Runs full forensic triage on a claim image.
- `verisight_active_clahe`: Applies CLAHE contrast normalization in LAB color space.
- `verisight_sweep_jpeg_ghosts`: Sweeps multiple JPEG qualities to find spliced patches.
- `verisight_azimuthal_fourier_probe`: Frequency-domain analysis for diffusion grid patterns.
- `verisight_sensor_noise_prnu`: Extracts sensor noise and checks local PRNU variance.
- `verisight_perceptual_hash`: Computes a perceptual hash to find duplicate photos.
- `verisight_micro_roi_seam_check`: Checks boundary seams for discontinuity.

## FAQ

### How do I run it?

Use uvx verisight mcp, or add it to a Claude Desktop or Cursor config with command uvx and args verisight mcp.

### Does it need a GPU?

The README says it runs on CPU with no neural network weights.

### What verdicts does it return?

Examples in the README include AUTO_APPROVE_CLAIM, HUMAN_IN_THE_LOOP_REVIEW and REJECT_SYNTHETIC_TAMPER.

### Does it call external APIs?

The README describes local OpenCV analysis, with optional evidence storage on Amazon S3 and DynamoDB.

### Is VeriSight MCP free?

Yes. VeriSight MCP is open source, so you can use and self-host it for free.

### How do I add VeriSight MCP to Claude Code?

Run this command in your terminal: claude mcp add verisight -- uvx verisight

### What are the best alternatives to VeriSight MCP?

The closest alternatives to VeriSight MCP are REA, AiSOC, redamon, x64dbg-mcp-server and Agent Roast.

## Where this comes from

- Description: written from the project's README.
- Install steps: from the official MCP Registry entry.

## Alternatives

- [REA](https://ai-powered.com/mcp/rea): REA is an open-source MCP server for developer tools.
- [AiSOC](https://ai-powered.com/mcp/aisoc): AiSOC is an open-source MCP server for ai models and llm tooling.
- [redamon](https://ai-powered.com/mcp/redamon): redamon is an open-source MCP server for ai models and llm tooling.
- [x64dbg-mcp-server](https://ai-powered.com/mcp/x64dbg-mcp-server): x64dbg-mcp-server 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.
- [Cisa Kev](https://ai-powered.com/mcp/cisa-kev): Cisa Kev is a freemium MCP server for security.

---
Source: https://ai-powered.com/mcp/verisight-mcp
