You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
MCP server support — register stdio or HTTP MCP servers in Settings and expose their tools to the in-app Agent and bundled terminal CLI agents.
More agent providers — Ollama, LM Studio, Hugging Face, Cloudflare Workers AI, Together AI, and Portkey join OpenRouter, Anthropic, and OpenAI.
Workspace Kanban — plans now have a pinned Kanban board with plan/task lanes, drag-and-drop status changes, search, and Markdown write-back.
Agent-authored Mermaid diagrams — create diagrams from chat or plans, keep them beside .agents/plans/, and export to Markdown or PDF.
HeartBeat + Memory Indexer — background services keep generated memory notes fresh across open workspaces.
Notifications and status line — a titlebar bell, deep-link notifications, and a low-noise bottom status bar surface active work without opening panels.
New look and 32 themes — redesigned BLXCode theme, BLXCode Legacy, new light variants, Claude Code dark theme, global radius, and font controls.
Voice, editor, and terminal polish — push-to-talk, local Whisper model management, Vim mode, named terminals, session compaction, and context-window metrics.
New
Settings → MCP: add, edit, enable, disable, and test MCP servers; terminal CLIs get managed project-scoped MCP config files.
AI plans and tasks: generate Markdown plans or task lists from a short prompt in the Plans panel.
Kanban center tab: every workspace gets a pinned plan board while Terminals remain the default active view.
Diagram gallery: plan-linked Mermaid diagrams render in a centered gallery and can be saved as .md or .pdf.
Agent session stats: provider/model, context usage, turns, tool calls, subagents, and session cost are visible in the Agent panel.
Terminal view modes: switch between Grid, Canvas, and Swarm layouts for terminal-agent workspaces.
Custom titlebar: brand cluster, workspace breadcrumbs, Navigate menu, notifications, and Help/About live in one themed header.
Beta update channel: opt into prereleases from Settings and get background update notifications.
Improved
File preview is now CodeMirror 6 — read-only preview and edit mode share highlighting, folding, selection, and editor behavior.
Git graph now uses VS Code-style lanes, richer hover details, and safer GitHub origin links.
Memory moved to the center with a wider layout, split view, and dedicated Memory settings.
Agent terminology is now consistently BLXCode Agent across the UI and docs.
Rules, skills, and plans gained cleaner category filters, live search, and more consistent card styling.
Fixed
Plan deletion now removes linked Mermaid diagrams and manifests.
Inactive Memory center tabs no longer remain visible beside Terminals.
Push-to-talk errors now surface as localized toasts.
Git and filesystem commands run off the main thread in more places, reducing UI stalls.
Model-round line numbers no longer show hash-derived values.
The per-message TTS button only appears when voice is actually configured.
Good to know
MCP tool sets are fixed at session start; reset the Agent session after changing MCP servers and reload terminal CLIs after MCP config changes.
Ollama and LM Studio need no API key, only a reachable /v1 server URL.
Local push-to-talk requires a downloaded Whisper-compatible model file; default builds still support cloud PTT.
Named terminals change display names only; slot_id remains the technical identity.
Existing users keep their Memory/sidebar preferences; new users start with the Memory right panel off.