v0.47.0

Live swarm views and smarter first runs

Highlights

• Alt+N now opens a full live swarm page with an ownership tree, animated status, and detailed activity for the selected agent
• First-run onboarding can launch a bounded read-only architecture review of your most recent substantive coding project
• External Claude, Codex, Pi, OpenCode, and Cursor sessions resume faster with bounded provider-safe history and reuse of existing continuations

Improvements

• Swarm spawn rows stay compact and stable in chat while rapidly changing todos, tool progress, and output remain on the live page
• Sponsored tool discovery requires specific reporting context and blocks recognizable secrets or personal data before any request is sent
• Todo, batch, and remote todo results render more consistently as compact cards with clearer retry and feedback-loop details
• Terminal file drops attach directly without being misrouted through skill commands

Fixes

• Imported logins activate the correct provider route and strongest available model before onboarding validates readiness
• Per-model image support is honored so images are only sent to routes that accept them
• Remote client turns reach every live attachment instead of only one connected view
• The browser bridge fails fast when its state is stale instead of attempting actions through an invalid connection
• Release smoke tests wait for Windows assets to be published

Full changelog: v0.46.0...v0.47.0

v0.46.0

Clearer swarm hierarchies and steadier math rendering

Highlights

• Spawn cards now show nested agents and their active work directly in the chat transcript
• Swarm prompts can be opened from inline controls without leaving the current workflow
• Multiline LaTeX stays stable while responses stream and standalone equations render as complete math blocks

Improvements

• LaTeX images scale to the terminal font size for more consistent visual proportions
• OpenAI OAuth usage displays adapt to dynamically reported context and rate-limit windows

Fixes

• Mouse capture remains enabled after restarting the TUI
• Task graph seeding rejects collisions safely instead of partially applying conflicting nodes
• Structured Markdown layouts preserve constrained widths when tables, lists, and blockquotes contain math

Full changelog: v0.45.0...v0.46.0

Sharper planning and richer terminal workflows

Sharper planning and richer terminal workflows

Highlights

• Image files can be dragged directly into the composer and attach immediately
• LaTeX can render as terminal images by default, including reliable multiline output
• Todo and swarm activity now use compact borderless cards, clearer notifications, and denser progress indicators

Improvements

• Todo goals now capture concrete feedback loops and enforce stronger measurable quality and end-to-end ownership gates
• Terminal titles stay synchronized with todo session titles while preserving legacy title fallbacks
• Startup avoids blocking terminal capability probes
• Managed phone-server deployment guidance and defaults use stronger AWS access controls

Fixes

• Agentgrep outline searches no longer duplicate relative file paths
• Exact fuzzy-search matches rank correctly
• Swarm cards resolve correctly outside gallery-filtered views
• Dropped images bypass slash-command routing and attach without delay

Full changelog: v0.44.0...v0.45.0

v0.44.0

More reliable swarms across platforms

Highlights

• Swarm sessions use broader collision-resistant identities and clean up finished terminal members more reliably
• Windows process handling now terminates detached task trees reliably and avoids startup and named-pipe failures
• Amazon Bedrock now accepts tools whose top-level JSON Schema uses oneOf, anyOf, or allOf

Improvements

• Swarm cards use calmer status indicators, stable worker identities, cleaner layouts, and tighter placement beneath spawn calls
• Markdown and LaTeX rendering handles more notation forms and malformed input safely
• Usage displays show reset countdowns, clear stale values correctly, and hide billing details when account authentication is unavailable
• Project links and website integrations now use jcode.sh

Fixes

• Remote server sessions initialize with the client's actual working directory
• Command previews strip ANSI control sequences before rendering
• Reload handoffs no longer issue unwanted terminal queries
• Swarm gallery previews build correctly with current member metadata

Full changelog: v0.43.0...v0.44.0

v0.43.0

Terminal math and smoother swarm work

Highlights

• LaTeX now renders as readable terminal math, including common inline, display, fenced, and environment-based notation
• Swarm waits now run asynchronously so the coordinator stays responsive while agents finish
• The CLI can override the working directory used by remote sessions

Improvements

• Swarm agent cards appear directly beneath spawn calls with clearer beta labeling
• Tool discovery covers more integration categories
• Todo guidance now emphasizes measurable goals and evidence-backed completion confidence

Fixes

• Images with mismatched format labels are corrected before sending so providers no longer reject valid image payloads
• Idle TUI sessions release retained heap memory more effectively
• Jcode login directs users to the correct pricing path

Full changelog: v0.42.0...v0.43.0

Richer swarm status and resilient sessions

Richer swarm status and resilient sessions

Highlights

• Inline swarm agent cards now show richer live status, expanded details, and clearer running animations
• Fuzzy search now tolerates typos so models, sessions, and commands are easier to find
• Running sessions are now animated in the resume picker and preserved more reliably during refreshes

Improvements

• Memory status uses a quieter count-first display and stays hidden when memory is disabled
• Mermaid rendering now follows one consistent feature toggle across configuration surfaces
• Skill slash commands can include a prompt after the command instead of requiring a second message

Fixes

• Detached and reloaded sessions now recover more reliably without lifecycle races, stale cleanup, or lost follow-up work
• Model catalogs refresh correctly after login imports and session resumes
• Provider retries now respect server delay hints, and interrupted OpenAI websocket responses start with a clean chain
• TUI redraw and retry amplification has been reduced, while remote debug snapshots now capture the current frame

Full changelog: v0.41.0...v0.42.0

v0.41.0

Inline generated images and Gmail by default

Highlights

• Provider-generated images now render inline beside their transcript entries in the TUI
• Gmail is now available by default in the full tool profile and can still be hidden with the tools disabled list
• The new /swarm-prompt command opens the active swarm routing prompt for straightforward project or global customization

Improvements

• Kitty image transfers are compressed to reduce terminal rendering overhead
• Explicit swarm spawns now require short labels so parallel agents are easier to identify

Fixes

• Anthropic API connections now honor their selected route and accept Jcode tools whose schemas previously used unsupported top-level combinators
• Side-panel Mermaid diagrams no longer render duplicate images
• Long inline image labels stay within one row instead of disrupting transcript layout

Full changelog: v0.40.0...v0.41.0

v0.40.0

Parallel browser control, clearer goals, and fresher model sessions

Highlights

• Browser automation sessions can now bind to individual browser windows, allowing parallel agents to work without interfering with each other's tabs
• Todo widgets now show each goal's hill-climbability score or taste-driven status, making it easier to see whether progress has a measurable objective
• Remote runtimes can use stable display names, and new server sessions now honor the latest saved provider and model defaults instead of stale server startup settings

Improvements

• The model picker can be limited to configured provider families, presents reasoning-effort rows consistently, and uses clearer alignment and login guidance
• Claude Fable 5 now defaults to high reasoning effort
• Memory relevance and extraction now default to GPT-5.6 Luna with reasoning disabled for lower latency when OpenAI credentials are available
• Tool transcript rows now retain useful summaries for scheduling, skills, MCP, discovery, Gmail, browser, and other tools across compaction and session reloads
• The session picker derives useful names from active todo groups or goal objectives when no manual session title is set
• Project-local skills are isolated to the workspace where they are defined so sessions in other repositories do not inherit them
• The iOS app now has a production-ready icon and automated TestFlight delivery

Fixes

• OpenAI model catalog refreshes recover automatically when an expired token initially returns 401
• Revoked or unknown phone connection tokens are rejected during the handshake, and the iOS app stops reconnecting in a loop and prompts for re-pairing
• New sessions created by a long-running shared server no longer display or use an outdated provider and model selection

Full changelog: v0.39.0...v0.40.0

v0.39.0

Cursor provider restored, active sessions manager, and cross-desktop launch hotkeys

Highlights

• The Cursor provider works again: jcode now talks to Cursor's current agent service directly instead of a decommissioned endpoint that failed with "Update Required" errors
• New active sessions manager (/active, or Left on an empty prompt when enabled) shows your live sessions with 'working'/'ready' badges so you can see which ones need attention

Improvements

• Launch hotkey setup now covers GNOME, KDE Plasma, Cinnamon, MATE, XFCE, bspwm, Hyprland, sway, and i3 on Linux, plus config-driven hotkeys on Windows
• The active sessions view refreshes live presence every couple of seconds and labels the session you opened it from

Fixes

• Cursor chat no longer fails with resource-exhausted / payment-required errors caused by the retired ChatService endpoint

Full changelog: v0.38.0...v0.39.0

v0.38.0

Light terminal theme support and goal-level todo scoring

Highlights

• The TUI detects light terminal backgrounds and adapts its colors automatically, with a config override for manual control
• Todos now carry goal-level hill-climbability scores with measurable objectives, and the agent is nudged to reframe taste-driven goals honestly instead of inventing proxy metrics

Improvements

• New Anthropic models get optimistic reasoning-cap defaults instead of being pinned to conservative limits
• The todo card hotkey default moved from Alt+P to Alt+X to avoid conflicts

Fixes

• Reasoning effort is passed correctly for GPT-family models on OpenAI-compatible gateways
• Phantom OpenRouter routes that pointed at unavailable models are no longer offered

Full changelog: v0.37.0...v0.38.0