English Français العربية

SignsOfAI MCP

Community AI and agents

Explainable AI-writing detection and originality checking in English and Spanish, plus citation checks and hidden-character detection.

Source code Documentation

Install

The installation steps are in the documentation below. Documentation

Documentation

From the project's README on GitHub, refreshed weekly. Links open on GitHub.

Contents
  1. SignsOfAI.Mcp — Model Context Protocol server
  2. Tools
  3. Install
  4. Claude Desktop
  5. Server tools (optional)
  6. Protocol note
  7. MCP registry

SignsOfAI.Mcp — Model Context Protocol server

Exposes the Signs of AI Writing engine as MCP tools, so Claude Desktop (or any MCP client) can analyze text, compare documents for copying, browse the sign catalog, and — optionally — measure perplexity and find cross-language paraphrases.

Built on the official ModelContextProtocol SDK (stdio transport) and a project reference to SignsOfAI.Core.

Tools

ToolWhat it doesRuns
analyze_ai_writingScore 0–100 + verdict + findings (overused vocabulary, rhetorical crutches, syntactic tells, burstiness), each with a fix🖥️ offline
check_originalityCompares 2+ documents against each other → overlap % + the actual shared passages🖥️ offline
search_catalogSearches the catalog of AI-writing signs (EN/ES), filter by keyword / language / category🖥️ offline
extract_distinctive_phrasesDistinctive phrases + ready-made exact-phrase web-search links🖥️ offline
inspect_charactersCharacters typing cannot produce — zero-width, homoglyphs, hidden tag characters — with the exact line and column of each🖥️ offline
check_citationsWhere a document contradicts its own reference list: a source cited but never listed, one DOI on two works, a year that has not happened🖥️ offline
compare_to_baselineHow a piece sits against the same person's earlier work (Burrows's Delta) — measured on their variation, never a verdict about who wrote it🖥️ offline
write_reportThe whole analysis as a document a person can keep, forward, or take to a committee — with the tool's own error rate printed on it🖥️ offline
measure_predictabilityPerplexity — how predictable/generic a model finds the phrasing☁️ server
check_paraphraseReworded/translated copies via sentence embeddings (EmbeddingGemma)☁️ server

Eight of the ten run entirely on the machine — the text never leaves it. The two marked ☁️ send the text to the SignsOfAI server (their descriptions disclose this); see Server tools.

Install

It's on NuGet as SignsOfAI.Mcp, a .NET tool. You need the .NET 10 SDK; nothing else to build.

# Run it on demand — no install step:
dnx SignsOfAI.Mcp --yes

# …or install the `signsofai-mcp` command once:
dotnet tool install --global SignsOfAI.Mcp

To hack on it instead, run it straight from the repo: dotnet run --project src/SignsOfAI.Mcp.

Claude Desktop

Add one of these to claude_desktop_config.json (Windows: %APPDATA%\Claude\claude_desktop_config.json), then restart Claude Desktop.

{
  "mcpServers": {
    "signs-of-ai": {
      "command": "dnx",
      "args": ["SignsOfAI.Mcp", "--yes"]
    }
  }
}

Or, if you installed the global tool:

{
  "mcpServers": {
    "signs-of-ai": { "command": "signsofai-mcp" }
  }
}

Working from a clone instead? Point it at the built DLL:

{
  "mcpServers": {
    "signs-of-ai": {
      "command": "dotnet",
      "args": ["C:\\path\\to\\SignsofAI\\src\\SignsOfAI.Mcp\\bin\\Release\\net10.0\\SignsOfAI.Mcp.dll"]
    }
  }
}

Server tools (optional)

measure_predictability and check_paraphrase call the SignsOfAI API. By default they use the PeopleWorks-hosted endpoint; override it with the SIGNSOFAI_API_ENDPOINT environment variable:

{
  "mcpServers": {
    "signs-of-ai": {
      "command": "signsofai-mcp",
      "env": { "SIGNSOFAI_API_ENDPOINT": "https://your-server" }
    }
  }
}

Unlike the offline tools, these two send the text off the device to run the model — the same disclosed, opt-in behavior as the web app's Predictability and Paraphrase features. check_paraphrase also needs the embedding feature enabled on the server.

Protocol note

MCP speaks JSON-RPC over stdout, so this server logs everything to stderr — never write to stdout from a tool.

MCP registry

Listed in the official MCP registry under the name below. The registry reads that line out of this README to verify that whoever publishes the registry entry also owns the NuGet package, so don't remove or reword it — publishing would start failing.

mcp-name: io.github.peopleworks/signs-of-ai

Frequently asked questions

What is SignsOfAI MCP?

Explainable AI-writing detection and originality checking in English and Spanish, plus citation checks and hidden-character detection.

How do I install SignsOfAI MCP?

Follow the installation steps in its documentation, shown on this page and in its repository.

Is SignsOfAI MCP open source?

Yes. Its code is on GitHub under the MIT license.

Who maintains SignsOfAI MCP?

It is published by peopleworks. AI.tn lists it but does not maintain or endorse it.

From the community awesome-mcp-servers lists and the project's GitHub repository. Listing is not an endorsement: review a server's code and permissions before giving it access to your data.