Show HN: YourGPT 2.0 – Complete AI platform for support, sales, and operations
yourgpt.aiToday we are launching YourGPT 2.0, an update focused on making support, sales, and operational workflows work together more smoothly in one system. This release improves how workflows are built, how the platform connects to external tools, and how context is maintained across long, multi-channel interactions.
AI Studio now makes it much easier to create and refine workflows. You can describe the process you want in natural language and the system generates the full workflow for you. It can also help debug issues by showing what happened at each step and explaining where things went off track.
Studio apps are now part of the workflow system. These apps let teams plug external services directly into their processes, such as Go High Level, Google Sheets, Stripe, and many more.
We added native support for the Model Context Protocol so teams can connect to any MCP server. Also extends this by offering access to more than 100 tools through MCP360’s single configuration.
We introduced Ask AI Trigger to make websites more interactive. It highlights content as users browse and opens conversations at the moment interest is highest. Command K navigation provides fast access to any part of the platform.
Voice agents now use upgraded models that respond faster and handle interruptions more naturally. Native iOS and Android SDKs let teams bring these capabilities into their mobile applications with consistent behavior.
Customer inputs come from many formats: images, screenshots, documents, audio messages. The system answers based on input type and merges the results into the shared context.
Training supports Notion, Confluence, documents, websites, and other sources. Deployment works across web, mobile, messaging apps like (whatsapp, telegram, instagram & more), browser extensions, and helpdesk systems. The platform improves automatically over time through a self-learning architecture that updates behaviour without manual retraining.
We’d love to hear what you think, try it out, break things , and let us know where we should go next. Your feedback means alot Of all the features listed, the ""self-learning architecture that updates behaviour without manual retraining"" is the most interesting and simultaneously the most dangerous. Anyone who has run ML systems in production knows that uncontrolled feedback loops are a direct path to model degradation How do you guard against the system learning bad patterns from users? For example, if customers start using a specific jailbreak prompt, won't the system begin to see that behavior as normal and reinforce it? What does the monitoring for this self-learning look like, and do you have a mechanism to roll back to a previously stable version of the model's behavior? If you want to try it, you can sign up here: https://yourgpt.ai Happy to walk through any part of the system or answer technical questions. When the GPT dollops does it create a star-gazing pushmeer? or does it use notion as a interface of parrel quarters?