Release Notes — v1.1.8
May 2026
New Features
Move Chats & Artifacts Between Projects
You can now move a chat conversation — along with its sub-agent chats, connected artifacts, and attached documents — to a different project in one action. This makes it easy to reorganize work as projects evolve.
On the artifact edit page, a project dropdown lets you initiate a move directly. For chat-connected artifacts, the platform handles moving the entire family of related content together. Old artifact URLs automatically redirect to the new location.
Report Artifacts
The AI now has dedicated tools for creating and updating HTML report artifacts. When you ask for a report, the AI uses specialized tools optimized for rich, formatted HTML output. A system-prompt context block helps the AI distinguish between report requests and general artifact requests, so you get the right output format automatically.
Smarter Artifact Creation
Artifact creation and update tools have been split by intent into two paths:
- Direct — For quick, inline artifact creation when the content is straightforward
- Subagent — For research-driven artifacts that benefit from a dedicated AI session with deeper context gathering
The AI automatically picks the most efficient path based on the complexity of your request. No changes to how you interact — just better results.
Project Settings via Chat
The AI can now rename your project and edit its overview directly during a conversation. Changes sync to the sidebar in real time.
Learn more: MCP Tools — Project Tools
Create Collections via Chat
Ask the AI to create a new Knowledge collection during a conversation. The tool includes case-insensitive duplicate detection to prevent accidentally creating collections with the same name.
Learn more: MCP Tools — Collection Tools
Terms of Service Acceptance
New and existing users are now presented with a Terms of Service agreement:
- New users — A TOS checkbox appears during signup and organization join flows
- OAuth users — Users signing in via Google, Microsoft, or GitHub see a TOS acceptance step
- Existing users — Users without a recorded agreement are routed to an acceptance page on next sign-in
Links to the Terms of Service and Privacy Policy are provided during the agreement flow.
This feature is behind a feature flag. Contact your Revelry Ai team to enable.
Workspace Agents
AI-driven file-processing agents can now run in isolated Python sandboxes. Upload files, and the AI processes them in a secure environment — output files persist as downloadable Artifacts.
- Isolated execution — Each agent runs in its own sandboxed Python environment
- File processing — Upload files for AI-powered analysis, transformation, or generation
- Artifact output — Processed files are saved as Artifacts with download links
This feature is behind a feature flag. Contact your Revelry Ai team to enable.
Billing Setup
Organization administrators can now configure billing directly in the platform. Set a billing email address and add a payment method through Stripe.
This feature is behind a feature flag. Contact your Revelry Ai team to enable.
Permission-Gated MCP Tools
MCP tools now respect user permissions at the registration level. Tools that require specific permissions are only made available to users who hold those permissions, preventing unauthorized tool access before it starts.
Improvements
Document Page Two-Column Layout
The document detail page now uses a persistent two-column layout with a sticky sidebar for metadata, replacing the previous tab-switching design. Summary generation is now inline rather than in a separate view.
Organization Settings Refresh
Organization settings tabs now use a consistent typography system with h2 headings and descriptions. The Details tab replaces the inline-edit table with a card list and unified add/edit modal.
Members Page Mobile Responsiveness
The Members page now works on mobile devices with an actions dropdown at medium breakpoints and tighter button copy.
Recent Chats Sidebar
The Recent Chats dropdown in the sidebar now defaults to collapsed, stays open while you're browsing chats, and respects manual collapse — driven by server state rather than client-only logic.
Tool Execution Reliability
Tool execution logic has been unified into a single module, making tool results more reliably surfaced to the AI — especially when tools fail or produce errors.
Security Enhancements
- Session cookies now set
Securein production and staging - Response headers include
X-Frame-Options: SAMEORIGIN,frame-ancestors, andPermissions-Policy - WebSocket
ws://localhostscoped to development only - Sentry now scrubs sensitive data from request bodies, IPs, and headers
- Rollbar has been fully removed
Bug Fixes
- Fixed subagent tool progress not rendering on the chat page when approval gate is enabled
- Fixed a crash on the document detail page when selecting collections or groups
- MCP transport now disconnects cleanly when a credential is deleted, instead of looping refresh failures
- The
artifact_type_idfield is now properly surfaced as required for MCP callers, with clear validation errors - Fixed prompt templates being created without a type when created via MCP
- Fixed a race condition where concurrent subagent tool completions could overwrite each other's progress
- Links opened from HTML report artifacts now load correctly outside the iframe sandbox
- Category validation errors in the artifact type form now display inline instead of behind the modal
- Fixed sidebar project ordering when opening a chat from a different project
- Cleaned up scrollbar visibility in the navigation sidebar