June 21, 2026 update 07:41 UTC

Clearer landing page, mobile visitor flow, and alpha labeling

The public first screen now shows a more product-like Lerix workflow with clearer model access, provider compatibility, and pricing cues.
Multi-model responses, Saved Agents, Codex auth and Relay, and chat storage choices are now represented with focused Lerix UI previews.
The mobile visitor experience now has a tighter account-first CTA flow, a compact menu, mobile-sized product previews, and visible alpha status in the top bar.

June 16, 2026 update 07:25 UTC

Mac status menu, Relay controls, update prompts, and favorite comparisons

Quick Composer is now available from the Mac status menu with its keyboard shortcut shown in the menu.
Codex Relay can now be turned on or off from the Mac status menu, with compact On/Off status while the desktop app is running.
Mac app windows and the status menu now behave more consistently when the app stays open in the background or fully quits.
Mac app update prompts now include restart actions in the status menu, including restart-after-download when an update is still downloading.
Saved Agent favorites can now be compared with favorite model configs while keeping Agent personality and file context attached to the Agent response only.
Multi-model responses now keep favorite comparison targets aligned with the visible composer state after selection and send.

OpenRouter, mobile web polish, and Mac app reliability

OpenRouter API keys can now be added as an AI service connection for Auto Router and compatible text chat models.
OpenRouter Auto Router responses now show the routed model in chat metadata when the AI service reports it.
Mobile web app controls now keep the composer and bottom sheets steadier on phone-sized screens.
Mac app account sign-in now completes account checks reliably in production desktop builds.
Mac app account and billing actions now handle desktop network restrictions more consistently.
Security improvements.

Account, billing, relay, comparison, support versioning, and trust updates

Account setup, sign-in, and billing flows now resume more consistently after refreshes, email confirmation, checkout, and account-session checks.
Encrypted account data access now gives clearer recovery-key and workspace-continuity handling during signup and sign-in.
Codex auth and Relay now handle same-account desktop and relay access more consistently where supported.
Multi-model responses now keep the selected comparison models visible in the composer after a send.
Support version details now reflect the tagged production release and build identity more consistently across Lerix surfaces.
Public docs now reflect current account, plan, storage, Codex, and model setup behavior with date-labeled product updates.
Security improvements.

Saved Agents, comparison controls, and Codex Relay polish

Pro accounts can now save Agents with a name, exact model config, personality prompt, and optional files from the model menu.
Agents can be starred as favorites, so favorite rows and shortcut switching preserve the Agent profile together with its model and access path.
Multi-model responses now show clearer per-response config details and keep branch selection stable when responses are included or dropped from the next round.
Codex auth and Relay now use clearer source labels and include reliability improvements.
Security improvements.

Direct streaming and cleaner response spacing

Streaming responses now render new text as it arrives from ChatGPT, Claude, Gemini, Grok, or a desktop relay instead of adding extra app-side pacing.
Prose and story responses now keep readable paragraph spacing when an AI service sends paragraph breaks as single newlines.

Smoother streaming across model providers

Streaming responses now use a shared provider-agnostic scheduler that smooths bursty model deltas for ChatGPT, Claude, Gemini, Grok, and desktop relay responses.
When a provider finishes sending a response, Lerix now renders any remaining buffered text immediately instead of continuing at the active streaming pace.

Streaming stability across model responses

Streaming responses now stay stable while Lerix saves and syncs chat updates, preventing partial assistant text from being replaced during an in-progress response.

Claude Opus 4.8, favorite-backed new chats, and model connection maintenance

Claude Opus 4.8 is now available for connected Claude API keys.
Favorite model config #1 now supplies the starting model setup for new chats when it is usable; otherwise Lerix falls back to the first connected model it can use.
Favorite model configs can now be edited from the Favorites panel and saved back with supported model controls before pressing Done.
Existing AI service connections can show newly supported API-key models without requiring you to recreate the connection.

Quick Composer, model favorites, new-chat defaults, and security improvements

The Mac app now includes Quick Composer for starting a prompt from a compact always-on-top composer window.
The model menu now supports up to 9 favorite model configs, with reordering, removal, and shortcut switching.
Settings now includes a new-chat default for choosing whether new chats start on this device or with encrypted sync when available.
New chats now open directly on the composer when a model is connected, so starting a fresh prompt takes fewer steps.
Security improvements.

Multi-model responses, support, model controls, and mobile web polish

The model menu now groups selectable models by AI service and keeps API-key or session access selection inside the active service.
ChatGPT model selection now keeps ChatGPT API-key access and Codex auth together under ChatGPT when both are available.
The composer now includes model search and favorite model configs, so the current model, access path, Intelligence, and Speed can become the starting setup for future new chats.
Pro accounts can compare multiple model responses from connected model configurations, review them side by side, and choose the response branch to continue.
Streaming responses can now be stopped from the composer, keeping the partial assistant message in the chat.
The latest user message can now be edited and resent. Edit mode keeps the original response visible until the resend starts and includes a cancel action.
Google API keys can now discover supported newly available Gemini text models automatically.
The model menu now exposes more AI-service-backed controls, including OpenAI response verbosity, Claude effort, Gemini thinking, and Grok reasoning where supported.
The mobile web app now uses larger chat controls, phone-native model and action sheets, and compact Settings tabs for easier use on phones.
Support and feedback now uses a signed-in form, with the reply email and account context attached from the Lerix account instead of an editable email field.

Initial release baseline

Initial released Lerix capabilities:

AI service API-key access is available for ChatGPT/OpenAI, Claude/Anthropic, Google/Gemini, and Grok/xAI.
No-account local use supports local-only API keys and chats without account sync.
Free accounts include encrypted API-key sync across signed-in devices.
Pro accounts include end-to-end encrypted workspace sync for chats, files, images, and workspace data, with 5 GB included storage and a 25 MB synced attachment limit.
The Mac app supports Codex auth, and Pro includes Codex Relay for same-account desktop workflows that support it.
The chat surface includes the composer, model menu, attachments, web browsing where available, image generation where supported, and Temporary Chat.
The app sidebar links to Docs and the product changelog.