Added
- Agent Sessions: opt in to discover, list, and focus live local or SSH-connected Codex and Claude Code sessions from the menu and CLI; discovery remains off by default.
- Wayfinder: add opt-in local gateway health, routing, savings, and latency usage with configurable loopback URL support. Thanks @tcballard!
- Menu bar: add a "Show reset time when quota runs out" option that replaces exhausted Percent, Pace, and Both values with the countdown until reset, then restores the selected metric afterward (#2028, #2027). Thanks @brahimhamichan!
- Quota warnings: add opt-in predictive pace alerts for Codex and Claude session and weekly limits, with one alert per risk episode. Thanks @vincent-peng!
- Codex: add GPT-5.6 Sol, Terra, and Luna pricing, including long-context, Priority, cache-write, alias, and automatic Pi cache repricing when rates change (#2023). Thanks @0xSMW!
- Codex: show Spark quota rows, with a provider option to hide them without hiding credits or other extra usage (#2013). Thanks @intellectronica!
- Claude CLI: surface model-scoped weekly limits alongside all-model usage without duplicating matching web limits. Thanks @janpollak!
- Kimi K2: add a Usage Dashboard shortcut to the human-facing legacy credits page. Thanks @joeVenner!
- Documentation: add detailed setup and troubleshooting references for Azure OpenAI, Perplexity, Mistral, and Qoder. Thanks @kiranmagic7!
Changed
- Settings: reorganize General, Notifications, Menu Bar, Menu, Advanced, and About; consolidate related checkboxes into pickers; standardize labels; and adopt an edge-to-edge Golden Gate sidebar with a hairline separator.
Fixed
- Refresh: keep all-provider manual refresh responsive while forced cost, credit, and dashboard enrichment finishes in a serialized background tail, and keep fixed intervals anchored to scheduled ticks. Thanks @Yuxin-Qiao!
- Menus: stop completed provider cards and plan-utilization rows from remaining in “Refreshing…” while unrelated provider or token-cost work is still running. Thanks @Yuxin-Qiao!
- Menu: keep native hover highlights aligned by deferring geometry-changing open-menu rebuilds until the pointer leaves the row. Thanks @Zihao-Qi!
- Settings: keep visual-only preferences and provider reordering on cached UI paths instead of refreshing provider quotas, while preserving refreshes for data-affecting settings. Thanks @Zihao-Qi!
- Display settings: keep display mode, work days, multi-account layout, and cost summary selectors interactive on macOS 27. Thanks @jordanschwartz-js!
- Quota warnings: keep compact per-window threshold editors available whenever notifications or usage-bar markers use them, preserve inherited overrides, and save edits on focus loss, Return, or window close. Thanks @Zihao-Qi!
- CLI server: retain timed-out route and provider work until it actually exits, preventing repeated requests or config changes from stacking background fetches. Thanks @Yuxin-Qiao!
- Widgets: show token-cost rows with their own age when they lag a fresh quota snapshot, and retry fast token-scan failures without waiting out the hourly cache. Thanks @irresi!
- Codex accounts: isolate authenticated OAuth and browser-cookie requests from shared URL caches and cookie stores, preventing one account's cached quota or identity response from appearing under another account and triggering false reset alerts (#1987, #2019). Thanks @harjothkhara!
- Codex: avoid false session-reset celebrations from transient zero-usage samples until the reset boundary advances. Thanks @kiranmagic7!
- Claude OAuth: honor the app's never-prompt policy in the bundled CLI and defer stale cache cleanup without touching Keychain until access is re-enabled. Thanks @Yuxin-Qiao!
- Claude CLI: resolve explicit-year, yearless, and time-only reset timestamps against the exact quota-window calendar occurrence, preserving DST, leap-day, and future reset semantics. Thanks @fanwenlin!
- Token costs: coalesce bounded pricing-catalog refreshes when a newly observed model is still unpriced, preserving its exact usage until pricing arrives. Thanks @iam-brain!
- Cost history: keep model breakdown menus steady while hovering, preserve compact rows, and make overflowing histories scrollable. Thanks @iam-brain!
- Antigravity: recover CLI listening ports from Linux procfs when
lsofis unavailable, including process network namespaces. Thanks @junmo-kim! - Gemini: prefer Google's paid-tier plan label over generic Free, Workspace, or Paid fallbacks while preserving acronym casing in the CLI. Thanks @Yuxin-Qiao!
- Ollama: recognize current WorkOS AuthKit sessions, fall back from expired sign-in redirects, and validate API keys against an authenticated endpoint while preserving refresh cancellation. Thanks @joeVenner!
- Kimi K2: report missing, blank, or rejected API keys clearly and trim surrounding whitespace before requests. Thanks @joeVenner!
- MiMo: flag a stale local-fallback cache in the summary (e.g.
stale 34d) so a tracker that has not been refreshed byScripts/mimo-usage.pyis not misread as live usage. Thanks @LeoLin990405! - Catalan: complete current strings, align instructional voice, and enforce catalog parity. Thanks @pmontp19!