Fuel iX Copilots — Week 42 · October 12–18

This week was mostly about cost visibility and control, with new spending caps and usage tracking for admins and users. We also made a number of reliability improvements, including a hotfix for MCP and OneSource.

  • Per-user monthly spending caps: Super admins can now set a monthly AI spending limit for each user. They can set an organization-wide default and override it for specific users. Users inherit the default unless given a custom limit.
  • Organization-level cost caps: Super admins can now turn on and configure a monthly inference budget for the whole organization, for better cost control across teams.
  • Monthly usage tracking for users: You can now see your current month's AI usage and spending against your cost cap, including how much of the cap you've used and when it resets.
  • Organization spending visible to admins: Organization admins can now see month-to-date AI inference spending against the organization's cost cap in Edit Organization settings.
  • Consistent usage indicator colors: Usage bars in the user menu and user settings now use the same color thresholds as organization usage: green below 70%, amber from 70–89%, and red above 90%.
  • Fixed an error when clearing a user's budget cap: Removing a per-user inference budget cap no longer causes a server error.
  • Service account knowledge bases in the Knowledge Hub: Knowledge bases created by service accounts now appear in the Knowledge Hub, so you can see all the knowledge resources available across your organization.
  • Longer display names: Display names and usernames can now be up to 50 characters, up from 30.
  • Faster recovery during AI provider outages: When an AI model provider stops responding, the platform now fails quickly (within 3 minutes). This helps keep Fuel iX responsive for everyone during provider incidents.
  • Fixed OneSource resource access: Resolved an issue that prevented resources from being managed and accessed properly in OneSource integrations. This was a hotfix following the incident on Thursday, October 15.
  • Expired sign-in sessions are now cleaned up: Fixed an issue where expired authentication tokens stayed in the system. Sessions now clear them out properly.
  • Clearer sign-in error messages: Failed authentication now returns an accurate "401 Unauthorized" response instead of a generic 503 error, which makes sign-in problems easier to diagnose.