For coding agents

One's Vibe, inside your build.

When someone tells their coding agent “I want to build an online chess game — is there anything like it?”, the agent can ask One's Vibe. It returns live, tryable AI-built products similar to the idea — real examples to open, how each is built, and a usage signal — as JSON.

Add it as an MCP server

The one-line way — exposes one tool, search_similar_projects(query, limit), to Claude Code, Cursor, Codex and any Streamable-HTTP MCP client:

claude mcp add --transport http onesvibe https://onesvibe.app/api/mcp

Or call the API directly

No key, CORS-open, rate-limited to 30 requests per minute per IP:

curl "https://onesvibe.app/api/reference?q=online%20chess%20game&limit=6"

What comes back

Each result carries the card-safe facts — never private scores or security findings:

title / taglinewhat it is

urlthe live product — open it or offer it to try

screenshota cover image URL (edge-cached)

stack / ai_toolshow it's built — build references

maturity / open_sourceprototype vs shipped; source availability

tries / recent_triesdeduplicated usage signal

buildmaker-declared: time, team size

The rules

Evidence, not a verdict. It tells you what already exists, how it's built, and whether it's live. It never says an idea is good or bad — the maker decides.

Try counts are usage, not quality. tries is a deduplicated, bot-filtered click signal. Low tries can mean “new,” not “bad.”

Web apps you can try in a browser. One's Vibe indexes browser-tryable products — not native/mobile apps or CLIs. No results can mean the idea lives in a different medium, not that it's undone.

See also the machine-readable llms.txt, or browse the gallery /explore.