Prospector Studio Release 2026.09.25 — SSO, Roles & Reliability Improvements

Identity & Single Sign-On

Clearer sign-in experience for restricted and misconfigured accounts. A user who signs in via SSO with no assigned role used to see what looked like a stuck, timed-out login — indistinguishable from an outage. They now land on a clear page explaining exactly what's missing (no groups reported from your identity provider, groups reported but not mapped to a role, missing email, etc.), so it's obvious this is a setup issue, not a system failure. A related crash affecting SSO users with no email on file is also fixed.

OIDC setup is easier for administrators. Configuring a single sign-on identity provider used to require manually constructing a technical redirect URL. It now displays automatically in the provider's detail panel.

Minor: A screen-reader label on the identity provider "Check connection" button now accurately describes what it does.

Role & Permission Management

Faster, safer role setup across environments. Administrators can now clone a role or permission set in one click, and export/import a full set of roles and permissions as a file — useful for keeping staging and production environments consistent. Imports show a full preview of exactly what will change before anything is applied, so nothing is overwritten silently. (Rolling out gradually, tenant by tenant.)

Case Management

Six reliability fixes for case and alert handling, addressing issues reported by customers:

• Editing a case's input values no longer silently drops required fields, which had been causing recurring automated tasks on that case to fail.
• Bulk-assigning cases to a user now works correctly and reports which specific cases failed, instead of failing silently across the board.
• Case comments now correctly show the most recent activity instead of only the oldest, so newer analyst notes are no longer hidden from view.
• Labels matching an active filter are now visually highlighted, making it easier to spot the case you're looking for.
• The Workflow Engineer AI persona can now read case labels it's asked to reason about, instead of failing.
• Alerts discarded during processing are now identified using the same ID your external security tools use, so they can actually be cross-referenced.

AI Agents & Automation

More reliable workflow execution at scale. Fixed two backend issues that could cause workflows with many parallel branches or array-style processing steps to intermittently fail or leak memory over time — both invisible day-to-day, but capable of causing slowdowns or failures under heavy use.

Agent-to-agent requests no longer randomly fail. A bug causing roughly 1 in 10 valid requests between AI agents to fail with an internal error is fixed.

Agents can now save documents containing short code snippets. Agents writing documents were incorrectly blocked from saving content that included short inline code snippets (like a query string), which validation mistook for invalid formatting. This is fixed.

Platform Reliability

Improved reliability for engagement settings in high-availability deployments. Certain engagement settings changes could intermittently fail with an internal error instead of behaving correctly; this is now fixed.

Persona → Documents view no longer crashes for high-activity tenants. Customers with long conversation histories could hit a blank screen or a stuck load opening a persona's Documents tab. It now loads reliably, with a graceful error message in the rare case of a backend hiccup.

Minor

• The "Dashboard" item under System Administration is renamed to "Pinned Charts" for clarity (it was easily confused with the separate "Dashboards" feature).