Have something to say?
Show context window usage at the bottom of conversations
It would be very helpful to display how much of the model's context window is currently used, at the bottom of each conversation. (in app) Why it matters In long conversations, answer quality tends to degrade: earlier instructions get overlooked, details from the middle of the thread are lost, and the model may start contradicting itself. Today there is no way to know whether this is happening because the conversation is getting too long, or whether older messages are being truncated or summarized in the background. A visible indicator would let users decide when to start a fresh conversation instead of guessing. Suggested implementation A discreet progress bar or percentage below the input field (e.g. "42k / 200k tokens, 21%") Recalculated for the currently selected model, since context window sizes differ between models A color change or warning above a threshold (e.g. 75%) suggesting to start a new conversation Optionally, a clear notice when older messages have been truncated or summarized Bonus: a "summarize and continue in a new conversation" button to carry key context over This would improve answer reliability and give users more control over long working sessions.
Security Report Follow-Up — Additional Critical Customer PII Exposure Identified
I wanted to follow up on a couple of security issues I recently discovered and reported through your internal channels. I originally reported these issues on August 2 at 17:29 IST, directly to contacts within the company, including the CTO, CEO, and an employee email address. One of the reports was submitted with the subject: “Unauthorized member addition to other projects, send legitimate invitation emails, and disclose project members’ PII — I found your project and members’ emails through this security issue.” Since then, I’ve identified another, potentially more critical security issue that appears to expose customer PII. I’m reaching out to report this responsibly and want to make sure it reaches the appropriate security or Development team as soon as possible. I haven’t received a response or acknowledgement to my earlier reports, so I’m concerned they may have been missed or may not have reached the right team. I’m intentionally not sharing any vulnerability details, customer information, or proof publicly. My goal is to give the company a reasonable opportunity to investigate and address these issues responsibly. I’m happy to provide the technical details privately through the appropriate security channel and work with the team to help verify and resolve the issues. Hopefully, this reaches the right people. I appreciate your time and look forward to working with the security team on this.
Trigger "End of Chat" for Automations
I want a native trigger in the agent automations called "End of Chat" or "Conversation Summary" that can automatically trigger actions (like sending an email summary via SMTP) when a conversation is finished or closed. Currently, I have to use workarounds like keywords, but a dedicated trigger would be much more efficient for sending discussion summaries to clients.
Interactive forms and checkboxes in agent messages
I want to be able to create agents that can send interactive elements like checkboxes, radio buttons, or forms directly in the chat. This would allow for more structured data collection (like a survey or a multi-step onboarding) instead of just text-based questions.
AI Agent: computer desktop access from the desktop app
Possibility to access user computer and execute command, access outlook email, create document word, ect..