English | [繁體中文](docs/zh/README.zh-TW.md)
# idea-reality-mcp
**How to check if someone already built your app idea — automatically.**
idea-reality-mcp is an MCP server that scans GitHub, npm, PyPI, Hacker News, and Stack Overflow to check if your startup idea already exists. It returns a 0–100 reality score with evidence, trend detection, and pivot suggestions — so your AI agent can decide whether to build, pivot, or kill the idea before writing any code.
**When to use this:** You're about to start a new project and want to know if similar tools already exist, how competitive the space is, and whether the market is growing or declining.
> **Project status (August 2026):** Maintenance mode. The tool works, stays free & open source, and the hosted API remains up; bug reports are reviewed, but no new features are planned.
> **Not just checking — building it?** After a reality check, open your idea as a public project on **[AngelRun](https://angelrun.vercel.app/new?utm_source=idea-reality&utm_medium=readme&utm_campaign=demand-cta)** — ship updates, climb the season, and get seen by angels.
[](https://pypi.org/project/idea-reality-mcp/)
[](https://smithery.ai/server/idea-reality-mcp)
[](https://opensource.org/licenses/MIT)
[](https://github.com/mnemox-ai/idea-reality-mcp/actions/workflows/ci.yml)
[](https://github.com/mnemox-ai/idea-reality-mcp)
[](https://pepy.tech/project/idea-reality-mcp)
## How it works
1. **Describe your idea** in plain English — e.g. "a CLI tool that converts Figma designs to React components"
2. **idea_check scans 5 databases** in parallel (GitHub repos + stars, Hacker News discussions, npm/PyPI packages, Stack Overflow questions)
3. **Get a 0–100 reality score** with trend direction (accelerating/stable/declining), top competitors, and AI-generated pivot suggestions
## What you get
```
You: "AI code review tool"
idea_check →
├── reality_signal: 92/100
├── trend: accelerating ↗
├── market_momentum: 73/100
├── GitHub repos: 847 (45% created in last 6 months)
├── Top competitor: reviewdog (9,094 ⭐)
├── npm packages: 56
├── HN discussions: 254 (trending up)
└── Verdict: HIGH — market is accelerating, find a niche fast
```
One score. Six sources. Trend detection. Your agent decides what to do next.
Try it in your browser — no install
## Quick Start
```bash
# 1. Install
uvx idea-reality-mcp
# 2. Add to your agent
claude mcp add idea-reality -- uvx idea-reality-mcp # Claude Code
```
**3. Ask your agent:** *"Before I start building, check if this already exists: a CLI tool that converts Figma designs to React components"*
That's it. The agent calls `idea_check` and returns: reality_signal, top competitors, and pivot suggestions.
Other MCP clients
**Claude Desktop / Cursor** — add to config JSON:
```json
{
"mcpServers": {
"idea-reality": {
"command": "uvx",
"args": ["idea-reality-mcp"]
}
}
}
```
Config location: **macOS** `~/Library/Application Support/Claude/claude_desktop_config.json` · **Windows** `%APPDATA%\Claude\claude_desktop_config.json` · **Cursor** `.cursor/mcp.json`
**Smithery** (remote, no local install):
```bash
npx -y @smithery/cli install idea-reality-mcp --client claude
```
## Setup & Configuration
First-time guided setup:
```bash
idea-reality setup
```
This walks you through:
1. **Terms acceptance** — data collection policy and disclaimer
2. **Platform detection** — auto-detects Claude Desktop, Claude Code, Cursor, Windsurf, Cline
3. **Config generation** — prints the exact JSON snippet for your platform
4. **Health check** — verifies MCP server, tools, and scoring engine
### Platform Configs
```bash
idea-reality config # interactive menu
idea-reality config claude_code # auto-installs via CLI
idea-reality config cursor # prints Cursor config
idea-reality config raw_json # generic MCP JSON
```
Supported: Claude Desktop · Claude Code · Cursor · Windsurf · Cline · Smithery · Docker
### Health Check
```bash
idea-reality doctor # core checks (~2s)
idea-reality doctor --full # + GitHub API, all 6 sources, Anthropic API
```
## Usage
**MCP tool call** (any MCP-compatible agent):
```json
{
"tool": "idea_check",
"arguments": {
"idea_text": "a CLI tool that converts Figma designs to React components",
"depth": "deep"
}
}
```
**REST API** (no MCP required):
```bash
curl -X POST https://idea-reality-mcp.onrender.com/api/check \
-H "Content-Type: application/json" \
-d '{"idea_text": "AI code review tool", "depth": "quick"}'
```
**Python**:
```python
import httpx
resp = httpx.post("https://idea-reality-mcp.onrender.com/api/check", json={
"idea_text": "AI code review tool",
"depth": "deep"
})
print(resp.json()["reality_signal"]) # 0-100
```
Free. No API key required.
## Why not just Google it?
**Your AI agent never Googles anything before it starts building.** `idea_check` runs *inside* your agent — it triggers automatically whether you remember or not.
| | Google | ChatGPT | idea-reality-mcp |
|---|---|---|---|
| **Who runs it** | You, manually | You, manually | Your agent, automatically |
| **Output** | 10 blue links | "Sounds promising!" | Score 0-100 + evidence |
| **Sources** | Web pages | None (LLM) | GitHub + HN + npm + PyPI + PH + SO |
| **Price** | Free | Paywall | Free & open-source (MIT) |
## Modes
| Mode | Sources | Use case |
|------|---------|----------|
| **quick** (default) | GitHub + HN | Fast sanity check, < 3 seconds |
| **deep** | GitHub + HN + npm + PyPI + Stack Overflow | Full competitive scan |
Scoring weights
| Source | Quick | Deep |
|--------|-------|------|
| GitHub repos | 60% | 22% |
| GitHub stars | 20% | 9% |
| Hacker News | 20% | 14% |
| npm | — | 18% |
| PyPI | — | 13% |
| Stack Overflow | — | 10% |
If a source is unavailable, its weight is redistributed automatically — so the deep-mode
weights above are renormalised over the sources that actually answered.
> **Product Hunt was removed on 2026-07-17.** It had carried 14% of the deep-mode weight
> since launch and had never returned a single result: the adapter asked for
> `posts(search: $query)`, and Product Hunt's API has no text search on posts at all
> (`Field 'posts' doesn't accept argument 'search'`). Its weight is now redistributed to
> sources that answer. If you need it back, it needs a real search surface — not a token.
## Tool schema
### `idea_check`
| Parameter | Type | Required | Description |
|-------------|---------------------------|----------|--------------------------------------|
| `idea_text` | string | yes | Natural-language description of idea |
| `depth` | `"quick"` \| `"deep"` | no | `"quick"` = GitHub + HN (default). `"deep"` = all 6 sources |
Full output example
```json
{
"reality_signal": 72,
"duplicate_likelihood": "high",
"trend": "accelerating",
"sub_scores": { "market_momentum": 73 },
"evidence": [
{"source": "github", "type": "repo_count", "query": "...", "count": 342},
{"source": "github", "type": "max_stars", "query": "...", "count": 15000},
{"source": "hackernews", "type": "mention_count", "query": "...", "count": 18},
{"source": "npm", "type": "package_count", "query": "...", "count": 56},
{"source": "pypi", "type": "package_count", "query": "...", "count": 23},
{"source": "stackoverflow", "type": "question_count", "query": "...", "count": 120}
],
"top_similars": [
{"name": "user/repo", "url": "https://github.com/...", "stars": 15000, "description": "..."}
],
"pivot_hints": [
"High competition. Consider a niche differentiator...",
"The leading project may have gaps in..."
]
}
```
## CI: Auto-check on Pull Requests
Use [idea-check-action](https://github.com/mnemox-ai/idea-check-action) to validate feature proposals:
```yaml
name: Idea Reality Check
on:
issues:
types: [opened]
jobs:
check:
if: contains(github.event.issue.labels.*.name, 'proposal')
runs-on: ubuntu-latest
steps:
- uses: mnemox-ai/idea-check-action@v1
with:
idea: ${{ github.event.issue.title }}
github-token: ${{ secrets.GITHUB_TOKEN }}
```
## Optional config
```bash
export GITHUB_TOKEN=ghp_... # Higher GitHub API rate limits
```
`PRODUCTHUNT_TOKEN` no longer does anything — the source is disabled and ignores it.
Setting it used to be worse than useless: it un-skipped a source whose query the API
rejects, so it reported "0 competitors on Product Hunt" into 14% of the deep score.
**Auto-trigger:** Add one line to your `CLAUDE.md`, `.cursorrules`, or `.github/copilot-instructions.md`:
```
When starting a new project, use the idea_check MCP tool to check if similar projects already exist.
```
## Roadmap
- [x] **v0.1** — GitHub + HN search, basic scoring
- [x] **v0.2** — Deep mode (npm, PyPI, Product Hunt), keyword extraction
- [x] **v0.3** — 3-stage keyword pipeline, Chinese term mappings, LLM-powered search
- [x] **v0.4** — Score History, Agent Templates, GitHub Action
- [x] **v0.5** — Temporal signals, trend detection, market momentum
- [x] **v0.6** — Onboarding CLI (`idea-reality setup`, `config`, `doctor`)
## Star History
[](https://star-history.com/#mnemox-ai/idea-reality-mcp&Date)
## Found a blind spot?
If the tool missed obvious competitors or returned irrelevant results:
1. [Open an issue](https://github.com/mnemox-ai/idea-reality-mcp/issues/new?template=inaccurate-result.yml) with your idea text and the output
2. We'll improve the keyword extraction for your domain
## Contributing
See [CONTRIBUTING.md](CONTRIBUTING.md) ([繁體中文](docs/zh/CONTRIBUTING.zh-TW.md)).
## License
MIT — see [LICENSE](LICENSE)
Built by [Mnemox AI](https://mnemox.ai) · [dev@mnemox.ai](mailto:dev@mnemox.ai)