Product Update: Selora AI v0.4.0 and the Road to OpenClaw

Selora AI v0.4.0

Selora AI v0.4.0 , our AI butler for Home Assistant , is out and focused on hardening the alpha. This release squashes early bugs that surfaced when using GPT models to generate automations, particularly around triggers involving times and booleans – and tightens the overall reliability of the integration.

Here’s what changed:

• Dismissal deduplication & validation: duplicate dismissals are now caught, and services and entities tied to LLM suggestions are validated before being applied.
• Dynamic suggestion cap: the number of suggestions now scales with home size, preventing overload on smaller setups.
• Automation value normalization: LLM-generated automation values are standardized so triggers like times and booleans are handled consistently, regardless of the model used.
• Stale automation cleanup: outdated Selora automations are purged on startup, keeping the system clean.
• Less verbose chat: LLM responses in the web chat are now more concise and to the point.
• Panel state sync: improved synchronization between the panel and Home Assistant for automations that become unavailable.

Refreshed UI

We’ve completely refreshed the Selora AI interface to make it lighter and easier to get started with. The new design reduces visual noise and puts the focus on what matters: getting your automations running.

First MCP Server Iteration

This release also ships the first iteration of the MCP server , enabling external AI agents to connect directly to Home Assistant through Selora AI. This is the foundation for the agentic workflows we’ve been building toward , and it’s only going to get more capable from here.

One-Click Install for SeloraBox

If you’re a SeloraBox user, getting started with Selora AI is now a single click. Head to Connect, enable the feature, and you’re up and running. No HACS, no manual repository setup, just flip the switch.

What’s Next

Next week we’re working on three fronts:

• OAuth 2.0 for the MCP server: bringing proper authentication to the MCP integration so external agents connect securely.
• Selora AI Cloud gateway : a managed LLM backend that makes Selora AI a true turn-key solution. No need to bring your own API key or configure a model provider.
• Bug fixes: continuing to harden the integration based on alpha feedback.

On the Roadmap: ZeroClaw

We want Selora AI to be the best way to interact with your smart home, and while the built-in web chat is already there, we’re exploring more communication channels.

OpenClaw does this well, and with our MCP server, users can already connect their Claw directly to Selora AI to leverage existing channels like WhatsApp or Telegram. But let’s be honest: running your own Claw isn’t for everyone. There’s significant configuration involved, security risks to consider, and ongoing maintenance.

We’re currently exploring ZeroClaw , a lightweight Claw written in Rust that could be embedded directly into Selora AI. We won’t ship a full OpenClaw instance, which is too heavy and a security hazard if not properly maintained. ZeroClaw takes a different approach: tiny footprint, fast, and secure by design. We haven’t started this work yet, but our experiments will begin soon and we’ll share more as it takes shape.

Stay tuned.