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.
Claude Dashboard Converter
James Davidson · Converter Claude
Anthropic launched Claude Dashboards — refreshable dashboards that connect to MCP. This demo shows a dashboard built on Omni’s MCP connected to an internal Omni instance, exported from Claude directly into Omni, bringing over queries that leverage the semantic model for governed, progressively hardened analytics.
Smart Chart Label Rotation
Will Chase · Visualization
Demo of improved chart label behavior: labels now fit horizontally when space allows, auto-rotate 45 degrees when they don’t, wrap multi-word labels to multiple lines, and prompt users to switch to a horizontal bar chart when too many categories are used in a column chart.
Python in Apps
Will Chase · Apps AI Python
Proof of concept showing Python running inside Omni apps via Pyodide, a WebAssembly port of Python that executes in the browser. A K-means clustering algorithm reruns on input change, with the Python script visible in the code editor — demonstrating that apps can handle far more than JavaScript.
Git Integration V2
Cathy Lennon · Git
Progress on the V2 Git integration: removing automatic commit backs in favor of checks for invalid YAML, introducing a local-clone model with explicit push/pull buttons, handling merge conflicts via side-branch resolution in GitHub, and rethinking the branch-switching UI.
Omni Agent LLM Support
Zef Li · AI LLMs
Demo of expanded LLM and connection support: Grok 4.7 integration, Snowflake key-pair authentication, Databricks Genie machine-to-machine OAuth, multiple LLM options in OmniAgent (GPT-6 Astra, Fable, Gemini), and regional availability toggles for EU customers.
Option to Prefer Apps Over Dashboards
Meredith Dodge · Apps
Apps are being promoted as the default option in the UI. The save dialogue now provides more context for choosing apps over dashboards, and a new app-preference toggle (defaulting to on) makes apps the default for new workbooks and for Blobby.
Delivery Ownership Transfer
Meredith Dodge · Scheduler Administration
New ability to search for a user and transfer delivery ownership — either one at a time or in bulk. When revoking a user, the system pre-selects their deliveries for transfer and provides feedback if the recipient lacks access to the underlying document.
Delivery Trash Restore
Corey Ruderman · Scheduler Administration
Feature-flagged functionality to send deliveries to trash and restore them. Users can undo from the dashboard or restore from a trash view in settings, with integration into the ownership-transfer workflow when a user has lost document permissions.
Permissions Improvements
Luke Bowerman · Administration
A collection of permission paper cuts: the ability to disable personal content sharing, clearer messaging when a user’s connection role prevents workbook access, share-dialog warnings when recipients lack connection access, and request-access explanations when viewer access is unavailable.
Dashboard Settings Relocation
Luke Bowerman · Dashboard Administration
Dashboard settings have moved into a new consolidated super-settings dialogue. An upcoming change will shift document and workbook model settings to the draft side (from publish) to support Git serialization.
Delivery Format Defaults
Luke Bowerman · Scheduler
New default format settings for deliveries and downloads at both organization and document level. Users can specify defaults (PNG, PDF, or Excel) instead of always defaulting to PNG, with per-document overrides available.
Playwright Optimizations
Luke Bowerman · Internal
Internal-facing Playwright test optimizations that have cut CI time by ~8%, with open PRs targeting 15%. Includes a nightly self-improving skill that evaluates Playwright tests and suggests consolidation, plus a PSA about closing stale PRs to fix GitHub merge-spinner delays.
Smart LLM Routing
Lei Tang · AI LLMs Efficiency
Concept demo for mid-turn LLM routing: instead of pre-determining request difficulty, the main agent explores first and detects when a request spans multiple topics, then prompts the user to switch to a more powerful model. Early eval results are more promising than random routing.
Custom Visualizations
Nathan Agrin · Visualization Dashboards
Update on AI-generated custom visualizations: bespoke chart components (bee swarm, 3D animated pie, Pac-Man pie), style-matching from pasted reference charts using ECharts, interactive editing, and the ability to add custom vis directly to dashboards. Coming to apps shortly.