Changelog

A running log of new features, improvements, and fixes in Cumulus.

May 2026

Collaboration

Multi-user chat sessions — Invite team members into any project chat for real-time collaborative AI sessions. Participants see each other’s presence and typing indicators live.
@-mentions — @mention a team member or the AI directly in a message. The AI responds fully when addressed, stays quiet during human-to-human exchanges, and uses a confidence assessment for ambiguous messages.
OWNER badge — The session creator is visually distinguished with an OWNER badge so it’s always clear who started the conversation.
AI Mute — Any participant can mute the AI mid-session, pausing its contributions without ending the conversation.
Unread indicators — Unread messages in shared sessions are highlighted so you don’t miss contributions from teammates.
Leave session — Participants can leave a shared session without deleting it for others.

Plans, billing & permissions

Teams and Pro plan tiers — Cumulus now has distinct Teams and Pro plans with feature gating. Capabilities are unlocked based on your organization’s active plan.
Three-tier membership — Organizations have three membership levels: OWNER, ADMIN, and MEMBER, each with appropriate access. Role management is available from Settings → Team.
Last-admin protection — The system prevents the last admin from being demoted or removed, ensuring organizations always have at least one admin.
Editor-tier permissions in projects — Project members can be granted Editor access, giving them the ability to manage sources, update project memory, and adjust project settings without full admin rights.
Stripe upgrade flow — The upgrade flow supports plan_id and coupon URL parameters, making it easy to link to specific plans or discounted offers.
Seat lifecycle management — Admins can manage seat assignments — inviting, removing, and reassigning members — with usage tracked against plan limits.

Chat experience

Inline chat rename — Rename any conversation or project-level chat directly in the sidebar without leaving the current view.
Move chat into a project — Standalone chats can be moved into a project workspace, keeping related conversations organized under a single project.
Fork a chat — Start a new session branched from any point in an existing conversation. The forked chat inherits the original thread’s source citations so the AI can continue referencing the same documents. → Learn more
Follow-up question toggle — Users can enable or disable the AI’s follow-up question prompts per their own preference from Settings → Personalization.
Model & reasoning controls in chat — Switch the underlying model and adjust reasoning depth directly from the chat input area without leaving the conversation.
Scheduled Tasks — Set up recurring AI tasks on a daily, weekly, biweekly, or monthly cadence. Results are delivered as in-app cards and email notifications. → Learn more

Integrations

Google Drive — full sync parity — Google Drive now supports bulk sync and delta sync on the same cadence as SharePoint. Folder browsing, personal Drive support, and webhook-driven updates are all included. → Learn more
SharePoint color-tag folder discovery — SharePoint connections can automatically discover folders marked with specific color tags in your tenant, making it easier to identify content to index without manual folder selection.
SharePoint folder include/exclude — Admins can specify which SharePoint folders to include or exclude from sync, giving precise control over what the AI can access.
Auto-renewing Graph webhook subscriptions — SharePoint integrations now automatically renew their Microsoft Graph webhook subscriptions before expiry, preventing silent sync gaps.

Knowledge & files

Least-privilege file access — A per-file toggle lets admins honour individual file ACLs at query time. When enabled, the AI only surfaces content from files the requesting user has permission to access.
Large-document progressive ingest UI — Uploading large documents now shows a step-by-step progress indicator: extract → chunk → embed → ready. You can see exactly where a document is in the pipeline.
Project Memories — Each project workspace now has a shared memory pool separate from personal AI Memory. Project Memories provide persistent context that all project members benefit from, without mixing into individual user memories.

Platform

Gemini embeddings — The ingest pipeline now supports Google Gemini as an embedding provider across document ingestion, improving retrieval quality for organizations using Gemini models.
Daemon upgraded to Python 3.13 — The background worker service has been updated to Python 3.13, picking up performance improvements and forward compatibility.

April 2026

Follow-Up Questions

The AI now asks clarifying questions before diving into complex requests. When your message could go in multiple directions — like “write a proposal” without specifying the audience or topic — the AI presents an interactive multi-question card with preset options and free-text input. Answer what you want, skip what you don’t, and the AI continues with your context baked in. → Learn more

Scheduled Tasks

Set up recurring AI tasks that run automatically on a daily, weekly, biweekly, or monthly cadence. Use the /schedule slash command to describe what you want, configure the timing, and the AI handles the rest — delivering results as in-app cards and email notifications. Manage, pause, or edit schedules from Settings or directly in chat. → Learn more

VERA Scoring System

Replaced the token-based VHR formula with VERA (Verified Effort Recovery Assessment) — a credible, auditable per-message scoring system. Admins can now see exactly how every VHR number is calculated, drill into individual user scores, and validate the methodology. → Learn more

Collaborative Chat

Multi-user AI chat within project workspaces. Add team members to any project chat, @mention users and AI, and see real-time presence and typing indicators. AI intelligently decides when to contribute — it responds fully when addressed directly, stays quiet during human-to-human exchanges, and uses a confidence assessment for ambiguous messages. File integrations (SharePoint, Google Drive) use an intersection model so AI only accesses files all members can see. → Learn more

March 2026

Web Search

Search the web for real-time information directly from chat. Use the /websearch slash command or just ask a question that needs current data. Results include inline citation chips linking to original sources. Queries are automatically sanitized to strip personal information. Admins can enable or disable web search from Settings. → Learn more

Slash Commands

Type / in the chat input to access quick commands: /websearch for web search, /image for image generation, and /data for document search. Navigate with arrow keys or Tab, select with Enter. Commands appear as a blue prefix in the input field. → Learn more

Image Generation Admin Toggle

Admins can now enable or disable image generation for their organization from Settings. When disabled, the /image command is hidden and the AI will not generate images. → Learn more

Tool Call Timeline

A new expandable timeline in chat shows each tool the AI uses in real time — with step-by-step progress, icons, and timing. Consecutive same-tool calls are deduplicated for a cleaner view.

Mobile Responsiveness

The app is now fully usable on mobile devices. Chat input stays visible (no more hiding behind the browser chrome), the sidebar opens as an overlay drawer with a hamburger menu, dashboard and settings navigation use horizontal tab bars, and the artifact panel opens as a full-screen overlay. Tool UI buttons are now touch-friendly on all screens.

Image Generation

Ask the AI to generate images directly in chat. Images render inline with a lightbox view and download button. Each user gets a weekly generation quota — usage is tracked per image. → Learn more

Knowledge Base

Upload documents (PDF, DOCX, XLSX, TXT, MD) to your organization’s knowledge base from Settings. The AI automatically searches indexed documents when answering questions. Includes file status tracking, chunk counts, and search/delete management. → Learn more

AI Personality

Admins can now shape the AI’s tone and perspective for their organization. Configure two personality traits, a brand worldview, value proposition, and custom instructions — with a live preview that updates as you type. → Learn more

Workspaces

Project folders with shared AI context are now live. Each workspace has five tabs: Overview, Chats, Sources (file uploads with status tracking), Team (role-based access), and Memory (persistent project context). Create workspaces from the sidebar, upload documents, invite collaborators, and compare files. → Learn more

Reasoning Levels

Control how deeply the AI thinks before responding. Four levels available via the settings panel in chat: Minimal (fast), Low (some reasoning), Medium (balanced), and High (deep analysis). → Learn more

Dictate

Voice-to-text input is now live. Click the mic icon in the chat input to start dictating — your words appear in the text field as you speak. Edit before sending, click again to stop. → Learn more

Citations

AI responses that draw from your internal data now include citation chips showing exactly which document and section was used. Click any citation to expand and read the source excerpt. → Learn more

Dashboards

User and admin dashboards showing usage frequency, estimated time saved, common topics, and adoption rates. Admins get an organization-wide view with per-user breakdowns. → Learn more

Self-Destructing Chats

Admin-controlled private chat mode is now available. Conversations are not saved to history unless the user chooses to save them. Enable from Settings → Organization. → Learn more

Member Invitations

Admins can invite team members by email. Invitees receive a branded email and land on an invitation page showing the org name and who invited them. New users go through onboarding automatically after accepting. Pending invitations can be resent or revoked. → Learn more

Google Drive Per-User Permissions & Personal Drives

Google Drive search results are filtered by each user’s actual Google Drive permissions. Admins can enable a “Personal Drives” toggle for team members. Sync now uses webhooks for faster content updates. → Learn more

SharePoint Sync Progress

SharePoint connections now show real-time sync progress with file-level detail — including pending, processing, indexed, and failed counts. Non-admin members see connection status without exposing configuration details. → Learn more

AI Memory

Import your AI context from ChatGPT, Claude, Gemini, or any other assistant. Memories are organized into three pools (identity, instructions, knowledge) with smart retrieval. Includes admin controls per org and per workspace. → Learn more

Chat UI Improvements

Inline mid-stream thinking indicators, full tool label coverage for all backend tools, and provisional chat sessions that update the sidebar instantly on send.

Super Admin Dashboard

Added /super-admin route gated to @suprvisr.ai email addresses
Overview: 9 KPI cards — total tenants, users, DAU, messages today, active subscriptions with MRR, SharePoint files, uploads, conversations, and total storage
Trends: 4 time-series charts (DAU, messages/day, new tenants/week, SharePoint files synced) with 7/30/60/90-day time range selector
Tenants table: Sortable view of all orgs with per-tenant 30-day stats, subscription status, and MRR
Re-engagement panel: Lists dormant tenants and users (inactive 7+ days or never used) with one-click “Copy email” for outreach
Mailing panel: Listmonk subscriber count, list breakdown, and recent campaign stats (configure via env vars)

February 2026

Dice Prompt Feature

Added a dice icon to the empty chat state
Clicking the dice inserts a role-specific example prompt into the input field
Prompts curated per role: Developer, Designer/Marketer, PM, Sales, Admin, and Other
Prompts stored in the database — updated without code deploys
Greeting in empty state now uses your preferred name from onboarding

User Onboarding Flow

Replaced modal-based onboarding with a full-page onboarding experience
Updated welcome message to “Hey there, welcome to Cumulus by Suprvisr AI”
Step 1 — Welcome: Animated greeting + preferred name input
Step 2 — Role Selection: Developer, Product Manager, Designer, or Other
Step 3 — Features Showcase: Tour of platform highlights
Name and role saved to membership profile
“Redo Onboarding” button added to Settings → Personalization

Help Center (This site!)

Launched docs.suprvisr.ai as the self-service help center
Full feature documentation for all platform capabilities
Getting started guides, prompt strategies, admin setup, and FAQ

Want to suggest what we build next? Use the Feature Board inside Cumulus to submit ideas and upvote your team’s requests.

FAQ