Demarker
Identify what an AI image carries (marks, metadata, watermarks), then clean it.
Install
Add this to your client's MCP configuration (Claude Desktop: claude_desktop_config.json; Cursor: .cursor/mcp.json; VS Code and others use the same "mcpServers" shape).
@demarker/mcpclaude mcp add demarker -e 'DEMARKER_API_KEY=<your demarker_api_key>' -e 'DEMARKER_MCP_MAX_CREDITS=<DEMARKER_MCP_MAX_CREDITS>' -- npx -y @demarker/mcp@0.1.2
{
"mcpServers": {
"demarker": {
"command": "npx",
"args": [
"-y",
"@demarker/mcp@0.1.2"
],
"env": {
"DEMARKER_API_KEY": "<your demarker_api_key>",
"DEMARKER_MCP_MAX_CREDITS": "<DEMARKER_MCP_MAX_CREDITS>"
}
}
}
}
Environment variables
DEMARKER_API_KEYsecret Optional. Needed to clean images and read balance. Create one at https://demarker.io/app/settings/apiDEMARKER_MCP_MAX_CREDITSOptional. Hard ceiling on credits one clean call may spend. Default 2.
Documentation
From the project's README on GitHub, refreshed weekly. Links open on GitHub.
Demarker agent tools
Open-source client, CLI, MCP server and agent skill for Demarker: see and remove what an AI generated image have hidden in them (visible marks, AI metadata, inferred hidden watermarks such as SynthID), then remove them!
Scanning is free and needs no account. Cleaning needs an API key and credits, and the person using it confirms a rights statement each time.
| Package | What it is |
|---|---|
demarker | CLI: npx demarker identify image.png |
@demarker/mcp | MCP server for Claude, Cursor, VS Code and others |
@demarker/client | Typed API client (no runtime dependencies) |
skills/demarker | Agent skill: npx skills add removeaiwatermark/demarker |
Quick start
npx demarker identify image.png # free, no key
export DEMARKER_API_KEY=sk_live_... # https://demarker.io/app/settings/api
npx demarker clean image.png --max-credits 1 --attest-rights
MCP, Claude Code: claude mcp add demarker -- npx -y @demarker/mcp. One-click Cursor and VS Code installs: https://demarker.io/developers
Safety by design
- Spending is capped:
--max-creditsis required, and the MCP server also enforces a per-call ceiling (default 2 credits). - The rights attestation is never defaulted. The CLI asks, or needs an explicit
--attest-rights; the MCP server asks the user through the client where supported. - Results never overwrite an existing file.
Development
pnpm install && pnpm build && pnpm test
This repository is published from Demarker's main codebase; issues and pull requests are welcome here. API reference: https://demarker.io/api/v1/openapi
MIT licensed.
Frequently asked questions
What is Demarker?
Identify what an AI image carries (marks, metadata, watermarks), then clean it.
How do I install Demarker?
It runs on your machine. In Claude Code: claude mcp add demarker -e 'DEMARKER_API_KEY=<your demarker_api_key>' -e 'DEMARKER_MCP_MAX_CREDITS=<DEMARKER_MCP_MAX_CREDITS>' -- npx -y @demarker/mcp@0.1.2. For other clients, paste the JSON configuration from this page into their MCP settings.
Does Demarker need an API key?
Yes. It expects DEMARKER_API_KEY, which you set in its configuration.
Is Demarker open source?
Yes. Its code is on GitHub under the MIT license.
Who maintains Demarker?
It is published by removeaiwatermark. AI.tn lists it but does not maintain or endorse it.
From the official MCP Registry. Listing is not an endorsement: review a server's code and permissions before giving it access to your data.