Stay updated with the latest features, improvements, and fixes
MCP v3.3
✓Added a "Validate a startup idea" command to the PreVibe MCP server. Run it from your AI app instead of remembering what to ask for: in Claude Desktop, click + in the message box, choose Connectors, then Add from PreVibe; in Claude Code, type /. Reconnect the server if you do not see it yet.
✓Your AI assistant now knows what PreVibe is for, so it can offer to validate an idea when you mention one, and suggest ideas to validate when you are still deciding what to build.
✓PreVibe is now listed in the Claude connectors directory as a Community connector. On any paid Claude plan you can find PreVibe under Settings -> Connectors and connect in one click - no server URL to paste. See the MCP Connector page for the steps.
MCP v3.1
✓Added Brand Names to the PreVibe MCP server: your AI app can now generate brand names for a research, list and read a set, suggest more, check domain availability, and delete - all without leaving the chat.
✓Added Brand Names, a contextual brand-name finder that reads your completed report and generates name ideas from your idea's positioning, category, and audience - so they fit your specific idea, not generic keywords. Each comes with an available domain, checked and priced. Just 1 credit.
✓PreVibe has a new home: previbe.io. previbe.app now redirects there, so your old links and bookmarks still work. Using the MCP server? Reconnect it from the connect page. Read why we moved
MCP v3.0
✓Simplified connecting the PreVibe MCP server to Claude, ChatGPT, and other AI apps. Just add the connector and sign in with Google - no more API keys to copy. The server now uses OAuth, and the endpoint is now /api/mcp.
✓Improved the accuracy of market data in research reports - figures that can't be verified are now clearly labeled
✓Improved the research chat: Shift+Enter now adds a new line, and the input grows as you type longer questions
✓The research chat now shows a live character count as you near the limit, and keeps your message if a send fails - so you never lose what you typed
✓Bigger, easier-to-read research chat (full screen on mobile) and a character count on the idea box
✓Research chat answers are now shorter and more to the point
✓Fixed reports occasionally mixing in another language - reports are now consistently written in English
✓Fixed a false 'insufficient credits' error that could appear when starting a research right after signing in
✓Fixed the credits counter in the header lagging behind - it now updates right after you spend or redeem credits, and when you come back to the tab
Version 2
✓Added a clear GO, PIVOT, or NO-GO verdict, a 0-100 viability score, and a short summary to every research report - no more praising every idea
✓Rebuilt research reports on live web research with clickable source links, not AI hallucinations
✓Added Deep Dive - a deeper analysis of your idea or a top opportunity, with Jobs-to-be-Done, pricing, risks, and a 90-day plan. Deep dives can be deleted from their page, just like researches
✓Cut the wait for a verdict to seconds - it now shows up before the full report finishes
✓Added credit packs and a new pricing page. New accounts start with 1 free credit, and purchased credits never expire
✓Made the research chat a floating bar so it's always at hand
✓Added free retries for failed reports and deep dives, and a running report no longer gets lost if you leave the page
✓Made research report and deep dive sections collapsible to keep the verdict on top
✓Expanded the idea input to accept larger, multi-line notes
✓Enabled zero data retention on all AI calls: providers delete your idea and report right after processing, and never train on them - read why this matters
✓Upgraded the PreVibe MCP server to v2 with deep-dive, free retry, and delete tools, plus shorter responses that save your agent's context
✓Launched the PreVibe Community - a members-only Discord for indie builders who validate before they build