Purpose
idea-reality-mcp is an MCP (Model Context Protocol) server positioned as a "pre-build reality check for AI coding agents." When you are about to start a new project, it can help you determine whether similar tools already exist, how intense the competition in that space is, and whether the market is growing or declining.
Main capabilities
- Describe your idea in natural language, for example "a CLI tool that converts Figma designs into React components."
- idea_check scans 5 databases in parallel: GitHub repositories and star counts, Hacker News discussions, npm/PyPI packages, Stack Overflow questions.
- Returns a reality score from 0-100, along with a trend direction (accelerating/stable/declining), major competitors, and AI-generated pivot suggestions.
- One score, six sources, trend detection, and your agent decides what to do next.
- If a source is unavailable, its weight is automatically redistributed, so the weights in deep mode are renormalized across the sources that actually return results.
- Supports automatic checking of feature proposals on Pull Requests via GitHub Action.
How to use
- Try it in the browser, no installation required.
- Ask your agent: "Before I start building, check whether this already exists: a CLI tool that converts Figma designs into React components." The agent will call idea_check and return reality_signal, major competitors, and pivot suggestions.
- Claude Desktop / Cursor: add to the configuration JSON. Configuration locations: macOS is ~/Library/Application Support/Claude/claude_desktop_config.json, Windows is %APPDATA%\Claude\claude_desktop_config.json, Cursor is .cursor/mcp.json.
- Smithery (remote, no local installation required).
- First-time guided setup: run idea-reality setup, which will guide you through terms acceptance(data collection policy and disclaimer), platform detection (automatically detects Claude Desktop, Claude Code, Cursor, Windsurf, Cline), configuration generation (prints the exact JSON snippet for your platform), health check (verifies the MCP server, tools, and scoring engine).
- Supported platforms: Claude Desktop, Claude Code, Cursor, Windsurf, Cline, Smithery, Docker.
- Any MCP-compatible agent can make MCP tool calls; a REST API is also available (no MCP required), free, no API key needed.
- Automatic triggering: add one line to CLAUDE.md, .cursorrules, or .github/copilot-instructions.md.
Use cases
- You are about to start a new project and want to know whether a similar tool already exists.
- You want to understand the level of competition in a certain space.
- You want to determine whether the market is growing or declining.
- Verify in a Pull Request whether a feature proposal duplicates an existing project.
Limitations and notes
- Project status (August 2026): maintenance mode. The tool is available, remains free and open source, and the hosted API stays online; bug reports will be reviewed, but no new features are planned.
- Product Hunt was removed on 2026-07-17. Since its release it had always accounted for 14% of the deep mode weight, but it never returned any results: the adapter requests posts(search: $query), while Product Hunt's API simply does not support text search on posts (Field 'posts' doesn't accept argument 'search'). Its weight has now been redistributed to sources that can return results.
- PRODUCTHUNT_TOKEN no longer works - that source is disabled and ignores it.
- Version history: v0.1 GitHub + HN search, basic scoring; v0.2 deep mode (npm, PyPI, Product Hunt), keyword extraction; v0.3 three-stage keyword pipeline, Chinese term mapping, LLM-driven search; v0.4 scoring history, agent templates, GitHub Action; v0.5 Time signals, trend detection, market momentum; v0.6 guided CLI (idea-reality setup, config, doctor).
- License: MIT. Built by Mnemox AI, contact email dev@mnemox.ai.
- If the tool misses an obvious competitor or returns irrelevant results, you can submit an issue with your idea text and output.
