Idea Reality MCP

What mnemox-ai/idea-reality-mcp MCP server does

The mnemox-ai/idea-reality-mcp MCP server gives coding agents a structured pre-build check for product ideas. An agent submits an idea in natural language, such as a description of a command-line tool, and receives evidence about whether related projects already exist.

The result includes a reality_signal from 0 to 100, a duplicate-likelihood assessment, trend direction, market-momentum sub-scores, source evidence, similar projects, and suggested pivots. Similar-project results can include repository names, URLs, descriptions, and star counts. The output is intended to help an agent decide whether to proceed, narrow the idea, or reconsider it before generating code.

How it works

idea_check searches public data sources in parallel. Quick checks use GitHub repository and star information plus Hacker News discussions. Deep checks also examine npm packages, PyPI packages, and Stack Overflow questions. The scoring system uses source-specific weights and redistributes a source's weight when that source is unavailable.

The tool accepts two arguments:

• idea_text is required and contains the natural-language idea to investigate.
• depth is optional and accepts quick or deep; quick is the default.

A quick request is intended for a fast sanity check, while deep mode provides a broader competitive scan. The documented examples show results such as repository counts, package counts, discussion counts, the most-starred similar project, trend direction, and a final verdict.

Setup and configuration

The mnemox-ai/idea-reality-mcp MCP server can run locally through uvx:

uvx idea-reality-mcp

For Claude Code, the README provides this registration command:

claude mcp add idea-reality -- uvx idea-reality-mcp

Claude Desktop and Cursor can use an MCP configuration with uvx as the command and idea-reality-mcp as its argument. The project also documents setup and configuration commands for platform detection, configuration generation, and health checks. Supported client or deployment targets listed in the README include Claude Desktop, Claude Code, Cursor, Windsurf, Cline, Smithery, and Docker.

A hosted REST endpoint is also documented for applications that do not use MCP. It accepts a POST request containing idea_text and an optional depth. The README states that the service is free and does not require an API key.

Tools and capabilities

The mnemox-ai/idea-reality-mcp MCP server exposes one MCP tool: idea_check. It can:

• Compare an idea with related GitHub repositories.
• Inspect Hacker News discussions for related activity.
• Count related npm and PyPI packages in deep mode.
• Count related Stack Overflow questions in deep mode.
• Produce a numeric reality signal and duplicate likelihood.
• Report accelerating, stable, or declining trend direction.
• Return similar projects and AI-generated pivot hints.

Limitations and notes

The results depend on the availability and quality of the public sources being queried. A missing source does not stop scoring; its weight is redistributed across sources that respond. Product Hunt was removed from the scoring pipeline on July 17, 2026, despite older descriptions mentioning it, because the documented adapter did not provide usable text search. The project is in maintenance mode: it remains free and open source, bug reports are reviewed, and no new features are planned.

The tool is a discovery and comparison aid, not proof that an idea is unique or commercially viable. Its evidence reflects searchable public activity, so private products and poorly indexed projects may not appear.