Ethics Check MCP

What r-huijts/ethics-check-mcp MCP server does

The r-huijts/ethics-check-mcp MCP server gives an AI assistant a structured way to question assumptions instead of only producing agreeable answers. It analyzes conversation context for issues across eight areas: privacy and data sensitivity, confirmation bias, transparency, fairness and discrimination, long-term consequences, stakeholder effects, power and manipulation, and alignment with ethical frameworks.

The server is intended for situations where an answer could reinforce a user’s existing view or overlook the effects on other people. It can prompt the assistant to ask counter-questions, present alternative perspectives, and examine whether a request is informative, persuasive, or potentially manipulative.

How it works

Four tools cover different parts of the analysis:

• Ethics Check reviews a conversation against the eight ethical dimensions and flags concerns such as privacy risks, bias, weak transparency, and fairness issues.
• Critical Thinking pushes for counterarguments, missing information, competing interpretations, and examination of the user’s assumptions.
• Ethics Guide frames difficult situations through utilitarian, deontological, and virtue-ethics perspectives.
• Ethics Learn remembers interactions and looks for recurring patterns in a user’s reasoning, including personal blind spots and biases.

The tools can operate from conversation context and can also be invoked directly. Example prompts include asking whether a position is intellectually honest, asking what a dissenting person might argue, or requesting analysis from several ethical perspectives.

Setup and configuration

The r-huijts/ethics-check-mcp MCP server is distributed as the npm package @r-huijts/ethics-vibe-check. A standard setup runs it with npx and passes the package name as the command argument:

{
  "mcpServers": {
    "ethics-vibe-check": {
      "command": "npx",
      "args": ["-y", "@r-huijts/ethics-vibe-check"]
    }
  }
}

Cursor has a one-click installation link and a manual MCP configuration example. Claude Desktop users can add the behavior through custom instructions or at the beginning of a conversation.

Custom instructions can identify the biases, assumptions, or difficult questions the user wants challenged.

Tools and capabilities

The server supports automatic background analysis and direct tool invocation. It is suited to reviewing product or engagement strategies, persuasive communications, research claims, and other decisions where confirmation bias, manipulation, stakeholder effects, or longer-term consequences deserve explicit examination.

Limitations and notes

The tool is designed to interrupt comfortable reasoning and may ask difficult or confrontational questions rather than provide the fastest agreeable response. Its effectiveness also depends on the quality of the conversation context and any custom instructions supplied by the user.