Product Update: Selora AI Panel Redesign, Proactive Suggestions, and MCP Server Launch
This week's highlights: Selora AI gets a full panel redesign with amber branding, a proactive suggestions pipeline surfaces AI-recommended automations, and the Selora MCP server ships its first release.
Selora AI: A Brand-New Look
The Selora AI panel has been completely redesigned with a fresh amber-branded look that feels polished and unmistakably Selora.
The layout has been reworked for clarity: automations are now displayed in a clean grid, with dedicated tabs for Flow, YAML, and History views. A new welcome screen (shown above) introduces Selora AI’s four core capabilities: creating automations in plain English, detecting patterns, managing devices, and answering questions about your setup.
We’ve also added bulk edit mode for managing multiple automations at once, Home Assistant Assist integration, and human-readable descriptions for triggers and actions, so you no longer need to decipher raw YAML to understand what an automation does.
Proactive Suggestions: AI That Comes to You
One of the biggest additions this week is the proactive suggestions pipeline. Selora AI now monitors your Home Assistant state history, detects usage patterns, and generates automation suggestions, surfacing them directly on both the dashboard card and the automations tab.
Behind the scenes, we built a full pattern detection engine with dedicated endpoints for browsing patterns and managing suggestions. The dashboard card has been redesigned with reliable toggles and SPA navigation, so you can review and act on suggestions without leaving the Home Assistant dashboard.
Hardened Security and Reliability
We tightened security across the board this week:
- WebSocket admin authorization: commands now require proper authentication
- LLM command injection prevention: input sanitization to protect against prompt injection attacks
- Webhook surface area cleanup: reduced the exposed API surface
- Automation pipeline fixes: non-string trigger values are now properly
coerced, and
initial_statefrom suggestion payloads is respected
These changes make Selora AI more robust for real-world installations where security is non-negotiable.
Selora MCP Server: First Release
The Selora MCP server has shipped its first release, complete with a full skill strategy rollout. This server enables Selora AI to expose its capabilities as structured tools, laying the groundwork for the agentic workflows we’re planning next.
Frigate Integration: On Hold
We’ve decided to deprioritize the Frigate integration for now in order to focus our efforts on Selora AI. The managed Frigate service on SeloraBox remains on our roadmap, but shipping a polished AI experience takes priority this sprint.
What’s Ahead
Next week, we’re tackling two major fronts:
- Agentic workflows: enabling Selora AI to autonomously chain multi-step tasks and decision-making, turning it from a recommendation engine into a true home automation agent
- Device integration: deeper device discovery, pairing, and control directly through the Selora AI panel
- Automation panel UX refresh: compact cards, search/filter bar, and overflow menus to reduce visual noise
- Suggestion quality improvements: tracking dismissal context so rejected patterns aren’t re-surfaced, plus correlation vs. causality guardrails to reduce spurious recommendations
Selora AI is fully open source. You can follow development and track changes on GitLab . Advanced users can already install the pre-release via our HACS integration .
Stay tuned for more updates.