Stay in the loop! Subscribe to the demo RSS feed to be notified when we post new demos.Demos highlight what we are working on or experimenting with, but are not a guarantee of release. Let us know your thoughts at support@omni.co.
App Building System Prompt Improvements
Jared Hardy · Apps AI
Before and after of app prompt improvements tested across multiple AI models (Anthropic, OpenAI, DeepSeek). App prompts now look better out of the box with refined styling.
Inline App Editing
Jared Hardy · Apps AI
Inline app editing prototype behind a feature flag. Select elements, change colors, adjust text styling, and edit app components directly without going back and forth with Blobby.
App Inspect Flow Improvements
Jared Hardy · Apps AI
Combined markup and data tabs into a single inspect view. See the queries driving each section, dive into result sets, and identify regions built via JavaScript or hard-coded values.
Interactive App Edit Prototype
Jared Hardy · Apps AI
Prototype for an interactive editing flow where you can interact with the app while editing. Includes multi-select to select and update multiple elements at once.
Ask Queue Prototype
Jared Hardy · AI
Queue functionality to batch multiple changes and send them to Blobby all at once, instead of asking one by one or manually editing everything.
Blobby Pumpkin Carving
Jared Hardy · Fun
Halloween Easter egg in Blobby Builder — carve a pumpkin with customizable colors, stems, shapes, and candles. Save it to replace your Blobby avatar, and add it to a Stardew Valley-themed pumpkin patch shared across your organization.
Reducing App Context Size
Dan Mosora · Apps AI
Feature flag that collapses pieces of app HTML based on IDs, JavaScript function names, and selectors to reduce context size for apps. Demoed with a Progress Quest RPG app requiring only two expanded sections to make a change.
Broken Query Detection
Dan Mosora · Apps AI
System now detects when a query is failing inside an app and can automatically fix it. Demoed with a Blobby Quest app where the boss data query was broken and the system identified and repaired it.
Better Embedded Blobby
Dylan Israel · AI Embed
Easier way to embed the Blobby chatbot for customers wanting to use the chatbot externally. Uses an auth flow with token creation, updated error reasons, and status codes. Supports a no-org-key approach using token and embed secret.
LLM Comparison MCP
Dylan Israel · MCP AI Embed
Look up available LLMs through MCP and pass them to a data request. Run comparisons to see how each model performs differently, with support for specifying thinking levels.
Usage Analytics via MCP
Dylan Israel · Analytics MCP
Access usage analytics for your organization directly without navigating to the analytics page. See active user counts and other usage metrics. Behind a feature flag, currently in code review.
Forecast Call App
Isaac Hergott · Apps Use Cases
App combining data display with Writeback functionality. Choose your name, make your quarterly call on deals, toggle Salesforce stage-five-or-six deals, add off-platform deals, and save a snapshot that writes back to a CSV workbook for tracking forecast over time.
AI Content Referencing
James Bueche · AI
Text editor improvements for AI and routine input. Typeahead references for skills, dashboards, apps, and routines. Paste a dashboard URL to deterministically run that dashboard’s queries with its defined filters instead of relying on the AI to guess.
AI Task List UX Improvement
Peter Schrader · AI UX
Upgrade to how tasks are displayed in the AI stream. When Blobby executes a complex multi-step plan, the task list follows as you scroll so you can always see what’s in progress and what was planned.
Verified Query Badges for AI
Peter Schrader · AI UX
Verified and unverified badges on queries to help admins understand if a query is safe to share. Flags whether a query uses custom fields or CTEs. Settings are evaluated at render time and retroactively updated when AI settings change. May change but want to signal quality through the AI exploration and query writing process.
AI Research Tagging UX
Peter Schrader · AI UX
Dynamically tag AI stream content as researched or not researched. Queries that aren’t part of the final answer are automatically folded, reducing screen clutter and making conclusions easier to follow.
Open Source Agent Package
Austin Aranda · AI API External Tools
Open source agent package with agents and sub-agents to help migrate off other tools and create polished apps from other LLMs. Demos include a Chargers calendar app with game highlights, a Whataburger food service map app, and revenue funnel visualizations — all with less prompting.
Context File Upload
Gustav Staprans · AI Context
Upload files (HTML, RTF, newsletters) to Omni and have it remember information at the organization and user level. Demoed with dog information — personal vet details saved to user context, general breed facts saved to org context. Converts non-markdown files automatically.
Slack Context Builder
Gustav Staprans · AI Slack Context
Automated context ingestion from Slack via the connectors hub. Omni studies Slack structure and saves channel organization to org memory and user-specific details to personal memory. Eliminates redundant external calls when asking about Slack channels.
Design.md Builder
Gustav Staprans · AI Design Embed
Create a design MD file with team styling info (role, name, hex codes) and have Omni automatically theme all future apps accordingly. Demoed with Arsenal colors producing a red-and-blue themed app. Could be used to drop an organization’s webpage as a brand template.
Dashboard Version History
Luke Bowerman · Dashboards Apps
History and settings shortcuts on dashboards and apps. Browse version history with detail on layout changes and filter configuration edits. State persists when switching between dashboard and workbook surfaces. Right-click context menus for quick access.
Admin Invite Improvements
Luke Bowerman · Administration
Invite admins directly from the invite dialog with bulk support. Adding more than one admin triggers a confirmation prompt. Part of broader authentication tweaks including the launch of external users.
External User Login Controls
Luke Bowerman · Administration
Granular login method controls for external and standard users. Turn off email login by default, enable it for all external users, or opt in individual standard users. Organization fallback preserves admin access if SAML breaks. All settings exposed via SCIM.
Dashboard Access Warnings
Luke Bowerman · Administration
Access warnings now visible on the dashboard viewing side, not just during editing. Choose opt-in, always on, or never on. Sharing flow discloses which users lack permissions and which queries may have access issues. New blue icon design for less alarming states.
Dashboard Saved Filter Sets
Luke Bowerman · Dashboards
Save collections of filter values as named filter sets on dashboards. Set defaults, share with others, and flip between saved configurations. Personal by default; editors can flag sets as visible to everyone. Coming next week.
Vega Legends Scrolling
Rob Miller · Visualization
Fixed Vega legends to scroll in the layers compiler, restoring parity with the old custom legends. Added styling knobs for appearance customization. On static downloads, legends are clipped with a count of hidden entries.
Custom Viz in Apps
Nathan Agrin · Visualization Apps
Custom visualizations using ECharts with interactive behavior in apps. Technical preview of shifting to dedicated custom components for reusability across the app framework. First look at custom viz running inside an app.
MCP/API Connectors in Apps
Nathan Agrin · MCP Apps API
MCP connectors integrated into the apps framework. Demoed with a live Notion MCP connector pulling real data into an app, plus a meta app that explores all available MCP tools. Respects read/write permissions from the connector setup. API endpoint integration planned.
Year-End Review App
Yesi Cisneros · Apps Use Cases
Spotify Wrapped-style year-end review app built for trials and embedded prospects. Features animated data reveals, background music, and a full report at the end. All data points are wired to live data with fake data for the demo.

