--- title: "Foxl v0.2.2: New Models, Faster" description: "Historical release notes for the first multi-provider model catalog, isolated browser profiles, message editing, and system reports." author: "Foxl Team" date_published: "2026-04-15" date_modified: "2026-07-16" canonical_url: "https://foxl.ai/blog/v0-2-2" markdown_url: "https://foxl.ai/blog/v0-2-2/index.md" image: "https://foxl.ai/blog/v0-2-2-model-catalog.png" social_image: "https://foxl.ai/blog/social/v0-2-2.png" content_type: "Release note" topics: ["Model catalog","Browser automation","Diagnostics"] products: ["Foxl Agent"] --- # Foxl v0.2.2: New Models, Faster > Non-Claude AI models on Bedrock, a dedicated browser profile that remembers your logins, message editing with attachments, and a one-click system report. - Author: Foxl Team - Published: 2026-04-15 - Last reviewed: 2026-07-16 - Reading time: 3 minutes - Canonical HTML: [https://foxl.ai/blog/v0-2-2]() ![The Foxl model picker from v0.2.2 showing the first multi-provider catalog and capability labels.]() Foxl v0.2.2 makes the app more useful day to day: more model choices, a browser profile that keeps its own login state, attachment-aware message editing, and a downloadable diagnostic report. ## One model catalog Foxl now registers models in one shared catalog instead of repeating model IDs and capabilities across five or more files. The release also adds GLM 5 from Z.AI and Kimi K2.5 from Moonshot AI on AWS Bedrock, routed through Bedrock's Converse API for tool use and streaming. The [current Foxl model catalog]() records which choices remain available today. Canonical model names also clean up usage reporting and provider-specific ID handling. The operational benefit is consistency: the selector, relay, routing, and usage records refer to the same catalog entry. > **Corrections (April 15 and April 24, 2026):** v0.2.3 fixed canonical model resolution and reasoning configuration for GLM 5 and Kimi K2.5. v0.2.19 fixed a separate issue where switching models mid-conversation on the foxl.ai provider left the cached agent on the previous model. On affected builds, the selector label alone did not prove that the requested model handled the turn. > **Availability update (June 17, 2026):** v0.3.18 removed GLM 5 and Kimi K2.5 from Foxl relay credits. They remain available in the desktop app when you connect your own AWS Bedrock account. ## Your browser, your logins Before this update, browser automation could share your personal Chrome profile. Agent activity could interfere with your tabs, and an already-running Chrome process could prevent the automation session from launching. Foxl now uses a dedicated local profile, as described in the current [browser automation guide](). Login state can persist across Foxl restarts without sharing cookies or tabs with personal Chrome. The profile is device-local, and websites can still expire sessions or require a fresh login. ## Edit messages with attachments You can now remove images or documents from a sent user message before regenerating it. Edit the message, remove the attachments you do not want, then choose Save & Regenerate. This is useful when an attachment was irrelevant or you want to retry with text only. Removing an attachment is itself a valid edit, so regeneration no longer requires a text change. ## One-click system report Settings now downloads one text report with app and OS details, agent state, recent conversation titles, excerpts from the latest conversation, audit entries, and filtered server logs. That detail makes the report useful, but it can also contain private text, file paths, tool names, and error messages. Review it before attaching it to a support request on Discord. ## Quality of life - The chat input focuses when you switch conversations. - Clicking Chat while already on the chat page starts a new conversation. - Conversation context menus are translated across all 10 supported languages. - Live Logs moved from the sidebar to Settings in this release. ## Upgrade and verify After updating, refresh the provider list before testing a newly added model. Run a short prompt with a tool call and confirm the model and provider in Usage, especially if you switch mid-conversation. To verify browser persistence, sign in through the dedicated automation profile, restart Foxl, and reopen the same site. Generate a System Report once so you know what it contains before you need support. Read the [full Foxl changelog]() or download the [Foxl v0.2.2 GitHub release](). ## References and further reading 1. [Current model catalog]() - Documentation 2. [Dedicated browser profile]() - Documentation