ClaudeClear MCP Server
# ClaudeRewrite MCP Server
MCP server for [ClaudeRewrite](https://removeclaudewatermark.com) - the #1 Claude watermark removal tool.
## Tools
### `rewrite_text`
Rewrites text to remove Claude's statistical watermark. Preserves meaning, facts, numbers, and code while breaking the token pattern detectors flag.
### `strip_image_metadata`
Removes C2PA "Made with Claude" credentials, EXIF, XMP, ICC, and all other metadata from images. Supports JPG, PNG, WebP, GIF, TIFF.
## Install
```bash
npx claudeclear-mcp
```
## Environment Variables
| Variable | Required | Description |
|----------|----------|-------------|
| `CLAUDECLEAR_API_KEY` | Yes (for rewrite) | Your API key from [removeclaudewatermark.com/account](https://removeclaudewatermark.com/account) |
| `CLAUDECLEAR_API_URL` | No | API base URL (default: `https://removeclaudewatermark.com`) |
| `CLAUDECLEAR_MCP_HTTP` | No | Set to `1` to run as HTTP server instead of stdio |
## Usage with Claude Code
Add to your Claude Code settings:
```json
{
"mcpServers": {
"claudeclear": {
"command": "npx",
"args": ["claudeclear-mcp"],
"env": {
"CLAUDECLEAR_API_KEY": "your-api-key"
}
}
}
}
```
## HTTP Mode
```bash
CLAUDECLEAR_MCP_HTTP=1 CLAUDECLEAR_API_KEY=your-key npx claudeclear-mcp
```
Server listens on port 3311 by default.
## Listed in
- [Claude Market MCP directory](https://www.claudemarket.ai/mcp)
## License
MIT
TDQS
Scored across 2 tools
The two tools are completely distinct: one operates on text content by rewriting, the other strips metadata from image files. There is no overlap or ambiguity in choosing between them.
Both tools follow the same verb_noun snake_case pattern (rewrite_text, strip_image_metadata), making the API predictable. The verbs clearly indicate the action and the nouns specify the target.
With only two tools the server is minimal, and while each tool serves a distinct purpose, the set feels thin for a general 'remove Claude traces' offering. It is acceptable for a narrow special-purpose server but is on the small side.
The server covers the two main content types that would carry Claude fingerprints: text and images. Minor gaps exist, such as document/PDF metadata cleaning or a verification/detection tool, but the core workflows are covered.