How to Check If Your Startup Idea Already Exists (Before You Build)

You have an idea for a developer tool. Before you spend weeks building it, you should check if someone already built it. This sounds obvious, but most developers skip this step — especially when using AI coding agents that start generating code immediately.

Y Combinator co-founder Paul Graham has a famous essay on the topic: founders often build solutions looking for problems. The fix is simple — search first, build second. Here are five methods, from manual to fully automated.

Five methods compared

GitHub Search

Low effort5–15 min

Coverage: Open source only

+ Free, shows star counts and recent activity
− Misses closed-source products, SaaS, and mobile apps. Keyword-dependent — you might miss repos that solve the same problem with different terminology.

Product Hunt Browse

Medium effort10–20 min

Coverage: Launched products

+ Shows real products with upvotes, launch dates, and maker profiles
− Only covers products that launched on PH. Many developer tools never launch there. Search is limited.

App Store / Play Store

Low effort5–10 min

Coverage: Mobile apps

+ Rating counts show market validation
− Irrelevant for developer tools, CLIs, APIs, or backend services.

Ask ChatGPT / Claude

Low effort1 min

Coverage: General knowledge

+ Fast, conversational
− No live data. LLMs hallucinate competitors that do not exist. Cannot tell you how many repos or packages are out there. No trend data.

idea-reality-mcp (automated)

None (agent runs it) effort3–15 sec

Coverage: GitHub + HN + npm + PyPI + Product Hunt + Stack Overflow

+ Scans 6 databases in parallel. Returns a 0–100 score with evidence, trend detection, and pivot suggestions. Runs inside your AI agent automatically.
− Focused on software/developer ideas. Less useful for physical products or niche B2B SaaS.

The real problem: your AI agent never searches

Manual searches work — if you remember to do them. The problem is that AI coding agents (Claude Code, Cursor, Copilot Workspace) start building the moment you describe your idea. They never pause to check if the idea already exists.

This is exactly what idea-reality-mcp solves. It is an MCP server — a tool your agent can call. Add one line to your agent config, and every new project starts with an automatic reality check:

Add to your CLAUDE.md or .cursorrules:

When starting a new project, use the idea_check MCP tool to check if similar projects already exist.

What a good result looks like

You: "AI code review tool"

idea_check →

├── reality_signal: 92/100

├── trend: accelerating ↗

├── GitHub repos: 847

├── Top competitor: reviewdog (9,094 ⭐)

├── npm packages: 56

└── Verdict: HIGH — find a niche fast

A score of 92 does not mean "do not build." It means the space is crowded — you need a specific angle. The pivot suggestions help you find that angle.

Try it now

Try idea-reality-mcp in your browser — free, no signup. Or install locally:

pip install idea-reality-mcp

Built by Mnemox AI. We build open-source MCP servers for trading memory and idea validation.