ZeroTrue MCP Server
Enables detection of AI-generated content in text, images, video, and audio via the ZeroTrue API, supporting multiple analysis tools and MCP-compatible clients.
README
ZeroTrue MCP Server
Official Model Context Protocol server for the ZeroTrue AI Detection API. Detect AI-generated text, images, video, and audio from any MCP-compatible agent or IDE.
Quick Start
Requires a ZeroTrue API key. Get one at zerotrue.app.
ZEROTRUE_API_KEY=zt_your_key npx -y @zerotrue/mcp stdio
Tools
| Tool | Description |
|---|---|
zerotrue_analyze_text |
Detect AI-generated content in plain text |
zerotrue_analyze_url |
Analyze media or text at a direct HTTP(S) URL |
zerotrue_analyze_local_file |
Analyze a local file by path (preferred for desktop/CLI clients) |
zerotrue_analyze_file |
Analyze a Base64-encoded file |
zerotrue_get_result |
Retrieve a previous analysis result by ID |
zerotrue_get_api_info |
Return API metadata and supported formats |
All tools accept optional isDeepScan and isPrivateScan flags where applicable.
zerotrue_analyze_local_file
The recommended tool for local MCP clients. Pass an absolute path: the server validates the file, detects its MIME type, and uploads it as a multipart form to ZeroTrue. No manual Base64 encoding needed.
{ "path": "/Users/alex/Downloads/photo.png", "isPrivateScan": true }
zerotrue_analyze_text
{ "text": "Paste the content here...", "isDeepScan": false }
zerotrue_analyze_url
{ "url": "https://example.com/video.mp4", "isPrivateScan": true }
zerotrue_get_result
{ "id": "246c6522-195d-45d3-af96-0f2360d2e0bc" }
Response Format
Every tool returns a consistent JSON envelope:
{
"ok": true,
"data": {
"id": "246c6522-195d-45d3-af96-0f2360d2e0bc",
"status": "completed",
"result": {
"ai_probability": 0.998,
"human_probability": 0.002,
"result_type": "ai_generated",
"feedback": "High probability of AI generation detected"
}
}
}
Errors use the same shape with ok: false:
{
"ok": false,
"error": {
"statusCode": 401,
"message": "ZeroTrue API key is required.",
"code": "ZEROTRUE_API_ERROR"
}
}
Client Setup
<details> <summary><strong>Codex</strong></summary>
[mcp_servers.zerotrue]
command = "npx"
args = ["-y", "@zerotrue/mcp", "stdio"]
[mcp_servers.zerotrue.env]
ZEROTRUE_API_KEY = "zt_your_key"
</details>
<details> <summary><strong>Claude Desktop / Claude Code</strong></summary>
Add to claude_desktop_config.json:
{
"mcpServers": {
"zerotrue": {
"command": "npx",
"args": ["-y", "@zerotrue/mcp", "stdio"],
"env": {
"ZEROTRUE_API_KEY": "zt_your_key"
}
}
}
}
</details>
<details> <summary><strong>GitHub Copilot CLI</strong></summary>
copilot mcp add zerotrue \
--transport stdio \
--env ZEROTRUE_API_KEY=zt_your_key \
--tools '*' \
--timeout 310000 \
-- npx -y @zerotrue/mcp stdio
</details>
<details> <summary><strong>VS Code</strong></summary>
Or add manually to your workspace .vscode/mcp.json:
{
"inputs": [
{
"id": "zerotrue-api-key",
"type": "promptString",
"description": "ZeroTrue API key",
"password": true
}
],
"servers": {
"zerotrue": {
"command": "npx",
"args": ["-y", "@zerotrue/mcp", "stdio"],
"env": {
"ZEROTRUE_API_KEY": "${input:zerotrue-api-key}"
}
}
}
}
</details>
<details> <summary><strong>Cursor</strong></summary>
Add to your Cursor MCP settings:
{
"mcpServers": {
"zerotrue": {
"command": "npx",
"args": ["-y", "@zerotrue/mcp", "stdio"],
"env": {
"ZEROTRUE_API_KEY": "zt_your_key"
}
}
}
}
</details>
<details> <summary><strong>JetBrains / IntelliJ</strong></summary>
Use the IDE MCP settings panel with a local stdio server:
{
"mcpServers": {
"zerotrue": {
"command": "npx",
"args": ["-y", "@zerotrue/mcp", "stdio"],
"env": {
"ZEROTRUE_API_KEY": "zt_your_key"
}
}
}
}
</details>
<details> <summary><strong>Remote / self-hosted (Streamable HTTP)</strong></summary>
Run the HTTP server:
ZEROTRUE_API_KEY=zt_your_key \
ZEROTRUE_MCP_PORT=8787 \
npx -y @zerotrue/mcp http
Or with Docker:
docker run -p 8787:8787 \
-e ZEROTRUE_API_KEY=zt_your_key \
zerotrue/mcp
MCP endpoint: http://localhost:8787/mcp
Health check: http://localhost:8787/healthz
Point your MCP client to the endpoint above. For production, place the server behind HTTPS and add authentication.
</details>
Configuration
| Variable | Default | Description |
|---|---|---|
ZEROTRUE_API_KEY |
(required) | ZeroTrue API key (zt_...). Required for all analysis tools. |
ZEROTRUE_API_BASE_URL |
https://api.zerotrue.app |
ZeroTrue API base URL |
ZEROTRUE_API_TIMEOUT_MS |
310000 |
Request timeout in milliseconds |
ZEROTRUE_MAX_FILE_BYTES |
104857600 |
Max local file size (default 100 MB) |
ZEROTRUE_MCP_TRANSPORT |
stdio |
Transport mode: stdio or http |
ZEROTRUE_MCP_HOST |
0.0.0.0 |
HTTP bind host |
ZEROTRUE_MCP_PORT |
8787 |
HTTP bind port |
ZEROTRUE_MCP_ENDPOINT |
/mcp |
HTTP MCP path |
Example Prompts
Is this text AI-generated? "The quantum entanglement of..."
Analyze https://example.com/profile.jpg with ZeroTrue and summarize the result.
Use ZeroTrue to check /Downloads/video.mp4. Keep the scan private.
Get ZeroTrue result 246c6522-195d-45d3-af96-0f2360d2e0bc and explain the verdict.
Security
- Store
ZEROTRUE_API_KEYin your MCP client config, never in source control. - Prefer
stdiomode for personal use: your key never leaves your machine. zerotrue_analyze_local_filecan only access files readable by the MCP server process.- For HTTP deployments, do not expose the
/mcpendpoint publicly without authentication and rate limiting.
Troubleshooting
ZeroTrue API key is required - Set ZEROTRUE_API_KEY in the environment where the MCP server starts.
File is not readable or does not exist - Use an absolute path and confirm the MCP server process has read access.
fetch failed - Check connectivity: curl https://api.zerotrue.app/api/v1/info. If details.cause.code is present, use it to diagnose DNS, TLS, or proxy issues.
Stale tool behavior after an update - Restart the MCP client. Most clients keep MCP subprocesses alive between tool calls.
Requirements
- Node.js
>= 20.11
License
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。