What's new in OpenTag

Fixed duplicate reports when routines or Conductor tasks completed work in the background. OpenTag now brings those results back to the original request before sending its final report, so you get a coordinated response when the work is ready.

Fixed large responses from connected apps being cut short and then treated as failed requests. OpenTag can now use more of a successful response within the supported size limit, avoiding unnecessary retries when working with large reports or lists.

Enable Meeting prep in Routines to receive private Slack briefings for upcoming meetings in your connected Google Calendar. OpenTag checks every 15 minutes for meetings starting within 45 minutes, using sources you can access to cover who you're meeting, your history together, open commitments, and talking points. Briefings link to supporting sources and note gaps in available background. Newly enabled routines prepare one initial briefing per meeting occurrence and stay quiet about later calendar changes; existing routines keep their saved instructions.

Billing now shows when a subscription payment has failed, when access will pause, and how an admin can pay the outstanding invoice. Paying before cancellation restores paid access automatically once no other confirmed overdue invoices remain. Fixed cases where an unrelated billing lookup error could prevent a confirmed payment from restoring access.

Ask OpenTag why a completed routine didn't send a report. It can now explain when the run found nothing new or followed an instruction to stay silent, using the reporting settings for that particular run. If the history doesn't establish a reason, OpenTag can say so instead of treating a missing report as proof of a delivery failure.

Fixed routine edits replacing instructions unrelated to your requested change. You can also ask for a one-time catch-up or a different reporting rule for a single run, such as sending a report even when there are no findings, while keeping the routine's saved instructions and recurring schedule intact.

Fixed OpenTag treating questions between teammates as requests for its help simply because it had replied earlier in the Slack thread. It now checks who a message is meant for more carefully, while still recognizing follow-up questions, corrections, and requests about its work without requiring another mention.

Fixed supported Stripe customer lookups returning masked names and email addresses needed for subscription and invoice reports. OpenTag can now use those details from your authorized Stripe connection to identify the customers behind the numbers. Older connections may need to be reconnected before the improvement is available.

OpenTag's updated logo now appears across sign-in, onboarding, the app sidebar, and supported app connection screens. Browser tabs, saved app icons, and link previews use the same branding, making OpenTag easier to recognize as you move between them.

OpenTag now starts some requests more quickly, reducing the initial pause before it begins working. This improves startup time; research and task completion times still depend on the work involved.

Fixed research stopping before it could pass its findings back to the rest of your request. OpenTag can now use the findings and sources from an interrupted research task while recognizing what remains unfinished, reducing the need to repeat that work.

Custom MCP setup now lets you choose OAuth, Token / API key, or No authentication for public servers. More servers that use WorkOS sign-in can now connect. If setup fails, OpenTag explains which step failed, including authorization or loading the server's tools, and tells you to check Integrations when it can't confirm whether the connection completed.

Fly.io

Connect Fly.io with an access token to inspect apps, machines, volumes, certificates, and Postgres clusters from OpenTag. Choose read-only access for checks or read and write access to manage resources within your token's permissions. The connection starts shared with your team; you can make it private.

Fixed temporary connection errors preventing requests and routines from finishing after their actions had succeeded. OpenTag can recover without repeating those actions. If it still cannot finish, the notice explains that earlier actions may have completed and identifies any confirmed Slack file delivery.

Ask OpenTag to find an earlier discussion in your current Slack DM, including messages in other threads. Search by a phrase or time range, then follow the source link back to the message. This works with the history available to OpenTag in that DM.

Fixed source links that could point to the wrong Slack channel or the start of a thread instead of the reply used in an answer. Answers in Slack and web conversations now link more reliably to the specific source message, including replies found through search.

Fixed Auto requests waiting indefinitely when a model provider stopped responding. OpenTag can now try a backup provider after a timeout or eligible failure, with another fallback if the first backup also fails. Canceling your request stops further attempts.