The catalog, over time

Catalog release history

Follow the app versions arriving in IntuneGet. Find an app, explore a month, and see what changed.

: 48,313
: 14,647
: 14,647

6 Sept 2026

Notepad4 (AVX2) zufuliu.notepad4.AVX2
zufuliu.notepad4.AVX2
16 Aug 2026
26.07r623426.08r6282
VirusTotal:
* Scintilla updated to 5.6.6, see https://www.scintilla.org/ScintillaHistory.html for the changes. * Localization for Portuguese (Brazil), contributed by @DAVIDVS-PR1MVS in PR #1289 and #1292. * Remove find/replace text length limitation, issue #1277. * Fill find replace edit box with last used string after restarting when "Settings -> Other Settings -> Remember Search Option" is enabled, issue #1282. * Allow NUL (var transform backslash for `\x00` or `\u0000`) in find/replace text. * Escape/unescape C characters now handle standard `\a`, `\b`, `\f`, `\n`, `\r`, `\t`, `\v` escape sequence, issue #1221. * Fix word wrap not show whole pasted long line, issue #1277, #1283. * Fix random crash caused by date race in multi-threaded word wrap. * Other bug fixes and improvements. Behavior Changes * Escape/unescape HTML/XML characters no longer map space (U+0020) to ` ` (U+00A0), tab (U+0009) to `&emsp;` (U+2003) and vice versa, as they are not equivalent. * Hex to character now strict on hexadecimal digits, only following forms (all treated as UTF-16/UTF-32 code point) with exact number of digits as `H` are supported: `\xHH`, `\xHHHH`, `\uHHHH`, `\UHHHHHHHH`, `U+HHHH` and `U+HHHHHH`. * Transform backslash now strict on hexadecimal digits, see table below. | Form | Description | |--|--| |`\xHH` | exact two digits, treated as raw byte in current code page (which different from Hex to Character but is same as built-in regex engine) | |`\uHHHH` | exact four digits, UTF-16 code point, could be surrogate pair |

Ghost Downloader

XiaoYouChR.GhostDownloader

26 Aug 2026

4.3.44.3.7

VirusTotal:

Tip Join the Ghost Downloader community on Discord: discord.gg/fKfhkPumEM Improvements 🔧 - CJK font selection now follows the app's language setting and supports Hong Kong/Macau character variants By @i0ntempest @XiaoYouChR - Reduced macOS app size by removing unused system framework dependencies By @XiaoYouChR Bug Fixes 🐛 - Fixed Cantonese (粤语) language option not loading its translation (#837) By @i0ntempest Tip 欢迎加入 Ghost Downloader Discord 社区: discord.gg/fKfhkPumEM 功能优化 🔧 - 中日韩字体选择现在跟随应用语言设置, 并支持香港/澳门字形 By @i0ntempest @XiaoYouChR - 移除未使用的 macOS 系统框架依赖, 减小应用体积 By @XiaoYouChR 问题修复 🐛 - 修复粤语语言选项无法加载翻译的问题 (#837) By @i0ntempest Downloads ────────────────────────┬───────────────┬───────────── Platform │Format │Link ────────────────────────┼───────────────┼───────────── 🪟 Windows │Setup.exe │Download ────────────────────────┼───────────────┼───────────── 🪟 Windows │Portable (.zip)│Download ────────────────────────┼───────────────┼───────────── 🍎 macOS (Apple Silicon)│.dmg │Download ────────────────────────┼───────────────┼───────────── 🍎 macOS (Intel) │.dmg │Download ────────────────────────┼───────────────┼───────────── 🐧 Linux │AppImage │Download ────────────────────────┼───────────────┼───────────── 🐧 Linux │.deb │Download ────────────────────────┼───────────────┼───────────── 🐧 Linux │.tar.xz │Download ────────────────────────┼───────────────┼───────────── 🤖 Android │.apk │Download ────────────────────────┴───────────────┴───────────── Full Changelog: v4.3.6...v4.3.7

Xiaomi.MiMoCode

2 Sept 2026

0.1.130.1.14

VirusTotal:

Summary Interrupted or errored turns stay recoverable: the incomplete assistant message is kept as a candidate, and /recover continues the turn append-only without rewriting history. Conversation titles can be generated through a dedicated API (POST /experimental/title, plus a genTitle helper) that falls back to a lite model when the preferred one is unavailable or returns invalid output. TUI voice input is upgraded to a tool-based voice_input protocol: ASR applies caret/selection-aware insert or set, submit stays explicit through the send field, and the call shape is validated so models can drive it reliably. Also in this release: compaction keeps a summary plus only the API rounds that arrive while that summary is generated, triggers at a flat 90% of the context window, and collapses the post-rebuild tail into an activity log; unused actor spawn/run resume arguments removed so models are pointed at send (already the supported path for waking and continuing a subagent); subagent actor use via exec limited to send; nested exec tool calls persisted and replayable in the TUI; same-session subagents inherit parent permission grants; per-session system prefix frozen in a snapshot for stable prompt-cache prefixes; optional auto_worktree notice (default off); skills updates for mate, pdf-official CJK fonts, and pptx-official image sourcing; plus provider/retry, MCP, compose-next, and build fixes. 中断或出错的回合仍可恢复：未完成的 assistant 消息会保留为候选，/recover 以 append-only 方式续跑，不改写历史。会话标题可通过专用 API 生成（POST /experimental/title 与 genTitle helper）；首选模型不可用或返回无效结果时，会回退到 lite model。 TUI 语音输入升级为基于工具的 voice_input 协议：ASR 按光标/选区做 insert 或 set，发送始终由 send 字段显式触发；调用 schema 经过校验，模型可以稳定驱动。本版本还包括：compaction 保留 summary 以及生成期间到达的 API 轮次，触发条件改为上下文窗口的固定 90%，重建后的 tail 收成活动日志；删除 actor spawn/run 中未实现的 resume 参数，把模型导向本来就能用的 send；子代理经 exec 使用 actor 时仅允许 send；嵌套 exec 调用会持久化且可在 TUI 回放；同会话子代理继承父级权限；会话 system prefix 冻结进快照以保持 prompt-cache 前缀稳定；可选的 auto_worktree 提示（默认关闭）；mate、pdf-official 中日文字体与 pptx-official 配图规则等 skill 更新；以及 provider/重试、MCP、compose-next 与构建相关修复。 Features - tui: upgrade voice input to the tool-based voice_input protocol (caret/selection insert/set, explicit send, validated schema) by @yanyihan-xiaomi in #2308, #2309 - session: project compaction as a summary plus compression-time tail; fire at a flat 90% of the context window; add MIMOCODE_COMPACTION_MAX_CONTEXT env default by @MiMoHardFather in #2294, #2263, #2260 - session: support append-only interrupted turn recovery (/recover, TUI recover action); keep errored assistants as recovery candidates by @wqymi in #2205, #2279 - session: collapse rebuild tail into a compact activity log instead of replaying verbatim tool results by @wqymi in #2275 - session: freeze the per-session system prefix in a persistent snapshot; move skill catalogs from user messages to the system tail; preserve prompt cache prefixes by @MiMoHardFather in #2289, #2287, #2258 - sdk: expose reliable POST /experimental/title / genTitle with lite-model fallback and write guards by @wqymi in #2228 - actor: drop unimplemented spawn/run actor_id resume arguments; keep send as the path for waking and continuing a subagent by @yanyihan-xiaomi in #2262 - actor: limit subagent actor calls through exec to send only by @wqymi in #2217, #2223 - exec: persist replayable nested tool parts (versioned snapshots, TUI renderers, debounce) by @yanyihan-xiaomi in #2222 - permission: same-session subagents inherit parent grants (fall back to current session id for inherit parent) by @wqymi in #2273 - config: add auto_worktree toggle (default off); path-based once-per-session main-worktree mutation notice by @wqymi in #2227, #2270, #2292 - skill: add mate builtin skill for desktop pet creation by @oscar-cao7 in #2225 - skill: pdf-official CJK font protocol (per-OS TTF ladder, CID STSong-Light fallback, never Helvetica) by @oscar-cao7 in #2189 - skill: pptx-official image sourcing decision rules (draw / cite / search / generate; limited image_gen budget) by @oscar-cao7 in #2265 - provider: route Xiaomi PTC through Responses custom exec; keep exec gateway available under GPT tool filtering by @MiMoHardFather in #2203, #2183 - tool-script: typed exec_command parameters (cmd, yield_time_ms, workdir); resolve MCP-prefixed tool aliases by @MiMoHardFather in #2180, #2202 Bug Fixes - provider: use 128K output defaults for large model families; align large-model output and compaction reserves by @MiMoHardFather in #2181, #2182 - provider: stabilize defaultModel fallback (cfg → recent → id-asc first allowed; no menu-priority sort) by @wqymi in #2306 - session: centralize provider retry coordination (terminal vs transient, Retry-After, no live-step replay after side effects) by @wqymi in #2216 - session: scope replace-agent base override to main/peer actors so subagents keep their own agent prompt by @MiMoHardFather in #2264 - session: allow relative paths within the workspace for file/search tools by @yanyihan-xiaomi in #2251 - session: defer text part creation until first delta (no empty text shells in DB) by @wqymi in #2193 - skill: replace broken skill-tools filtering with glob prefix matching; remove skill-search reminder infrastructure by @MiMoHardFather in #2166, #2168 - llm-server: advertise Server.listen so mimo llm-server issue resolves a real base_url; publish/unpublish on serve stop by @wqymi in #2293 - mcp: record local startup exit code and stderr tail on stdio failure by @wqymi in #2269 - mcp: show MiMoCode branding in OAuth callback pages and DCR client registration by @MiMoHardFather in #2301 - compose-next: require spec to land in the worktree (pipeline grill → workspace → spec → implement) by @yanyihan-xiaomi in #2307 - actor: scope fork contexts by session so repeated actor ids cannot collide across sessions by @wqymi in #2259 - build: prevent ancestor node_modules bun from hijacking compiled binaries (pin check + process.execPath) by @MiMoHardFather in #2291 - build-node: define MIMOCODE_VERSION for the node target so embedded builds stop extracting under builtin_skills/local/ by @yanyihan-xiaomi in #2297 Refactor - skill: move drive-mimo to repository scope (.mimocode/skills/) instead of the product-wide builtin bundle by @yanyihan-xiaomi in #2210 - session: simplify harness selection — explicit harness flag is the single source of truth; PTC transport no longer auto-selects the GPT toolset; MIMOCODE_CODEX_MODE=false forces the default harness for every model by @MiMoHardFather in #2243, #2252, #2303 - tool: remove standalone change_directory; require absolute working directories for shell/code-mode commands by @yanyihan-xiaomi in #2239 - prompt: rewrite Tools section as “Using your tools” guidance; instruction files default-on and decoupled from the dynamic system prompt flag by @MiMoHardFather in #2302, #2285 Internal / CI - session: retune the auto-overflow fixture for the flat 90% trigger by @yanyihan-xiaomi in #2266 - sdk: regenerate openapi and js sdk types (projection schema, harness description, compaction field docs) by @yanyihan-xiaomi in #2310 - chore: bump version to 0.1.14 by @yanyihan-xiaomi in #2311 Full Changelog: https://github.com/XiaomiMiMo/MiMo-Code/compare/v0.1.13...v0.1.14

MiMoCode (Baseline)

Xiaomi.MiMoCode.Baseline

2 Sept 2026

0.1.130.1.14

VirusTotal:

Summary Interrupted or errored turns stay recoverable: the incomplete assistant message is kept as a candidate, and /recover continues the turn append-only without rewriting history. Conversation titles can be generated through a dedicated API (POST /experimental/title, plus a genTitle helper) that falls back to a lite model when the preferred one is unavailable or returns invalid output. TUI voice input is upgraded to a tool-based voice_input protocol: ASR applies caret/selection-aware insert or set, submit stays explicit through the send field, and the call shape is validated so models can drive it reliably. Also in this release: compaction keeps a summary plus only the API rounds that arrive while that summary is generated, triggers at a flat 90% of the context window, and collapses the post-rebuild tail into an activity log; unused actor spawn/run resume arguments removed so models are pointed at send (already the supported path for waking and continuing a subagent); subagent actor use via exec limited to send; nested exec tool calls persisted and replayable in the TUI; same-session subagents inherit parent permission grants; per-session system prefix frozen in a snapshot for stable prompt-cache prefixes; optional auto_worktree notice (default off); skills updates for mate, pdf-official CJK fonts, and pptx-official image sourcing; plus provider/retry, MCP, compose-next, and build fixes. 中断或出错的回合仍可恢复：未完成的 assistant 消息会保留为候选，/recover 以 append-only 方式续跑，不改写历史。会话标题可通过专用 API 生成（POST /experimental/title 与 genTitle helper）；首选模型不可用或返回无效结果时，会回退到 lite model。 TUI 语音输入升级为基于工具的 voice_input 协议：ASR 按光标/选区做 insert 或 set，发送始终由 send 字段显式触发；调用 schema 经过校验，模型可以稳定驱动。本版本还包括：compaction 保留 summary 以及生成期间到达的 API 轮次，触发条件改为上下文窗口的固定 90%，重建后的 tail 收成活动日志；删除 actor spawn/run 中未实现的 resume 参数，把模型导向本来就能用的 send；子代理经 exec 使用 actor 时仅允许 send；嵌套 exec 调用会持久化且可在 TUI 回放；同会话子代理继承父级权限；会话 system prefix 冻结进快照以保持 prompt-cache 前缀稳定；可选的 auto_worktree 提示（默认关闭）；mate、pdf-official 中日文字体与 pptx-official 配图规则等 skill 更新；以及 provider/重试、MCP、compose-next 与构建相关修复。 Features - tui: upgrade voice input to the tool-based voice_input protocol (caret/selection insert/set, explicit send, validated schema) by @yanyihan-xiaomi in #2308, #2309 - session: project compaction as a summary plus compression-time tail; fire at a flat 90% of the context window; add MIMOCODE_COMPACTION_MAX_CONTEXT env default by @MiMoHardFather in #2294, #2263, #2260 - session: support append-only interrupted turn recovery (/recover, TUI recover action); keep errored assistants as recovery candidates by @wqymi in #2205, #2279 - session: collapse rebuild tail into a compact activity log instead of replaying verbatim tool results by @wqymi in #2275 - session: freeze the per-session system prefix in a persistent snapshot; move skill catalogs from user messages to the system tail; preserve prompt cache prefixes by @MiMoHardFather in #2289, #2287, #2258 - sdk: expose reliable POST /experimental/title / genTitle with lite-model fallback and write guards by @wqymi in #2228 - actor: drop unimplemented spawn/run actor_id resume arguments; keep send as the path for waking and continuing a subagent by @yanyihan-xiaomi in #2262 - actor: limit subagent actor calls through exec to send only by @wqymi in #2217, #2223 - exec: persist replayable nested tool parts (versioned snapshots, TUI renderers, debounce) by @yanyihan-xiaomi in #2222 - permission: same-session subagents inherit parent grants (fall back to current session id for inherit parent) by @wqymi in #2273 - config: add auto_worktree toggle (default off); path-based once-per-session main-worktree mutation notice by @wqymi in #2227, #2270, #2292 - skill: add mate builtin skill for desktop pet creation by @oscar-cao7 in #2225 - skill: pdf-official CJK font protocol (per-OS TTF ladder, CID STSong-Light fallback, never Helvetica) by @oscar-cao7 in #2189 - skill: pptx-official image sourcing decision rules (draw / cite / search / generate; limited image_gen budget) by @oscar-cao7 in #2265 - provider: route Xiaomi PTC through Responses custom exec; keep exec gateway available under GPT tool filtering by @MiMoHardFather in #2203, #2183 - tool-script: typed exec_command parameters (cmd, yield_time_ms, workdir); resolve MCP-prefixed tool aliases by @MiMoHardFather in #2180, #2202 Bug Fixes - provider: use 128K output defaults for large model families; align large-model output and compaction reserves by @MiMoHardFather in #2181, #2182 - provider: stabilize defaultModel fallback (cfg → recent → id-asc first allowed; no menu-priority sort) by @wqymi in #2306 - session: centralize provider retry coordination (terminal vs transient, Retry-After, no live-step replay after side effects) by @wqymi in #2216 - session: scope replace-agent base override to main/peer actors so subagents keep their own agent prompt by @MiMoHardFather in #2264 - session: allow relative paths within the workspace for file/search tools by @yanyihan-xiaomi in #2251 - session: defer text part creation until first delta (no empty text shells in DB) by @wqymi in #2193 - skill: replace broken skill-tools filtering with glob prefix matching; remove skill-search reminder infrastructure by @MiMoHardFather in #2166, #2168 - llm-server: advertise Server.listen so mimo llm-server issue resolves a real base_url; publish/unpublish on serve stop by @wqymi in #2293 - mcp: record local startup exit code and stderr tail on stdio failure by @wqymi in #2269 - mcp: show MiMoCode branding in OAuth callback pages and DCR client registration by @MiMoHardFather in #2301 - compose-next: require spec to land in the worktree (pipeline grill → workspace → spec → implement) by @yanyihan-xiaomi in #2307 - actor: scope fork contexts by session so repeated actor ids cannot collide across sessions by @wqymi in #2259 - build: prevent ancestor node_modules bun from hijacking compiled binaries (pin check + process.execPath) by @MiMoHardFather in #2291 - build-node: define MIMOCODE_VERSION for the node target so embedded builds stop extracting under builtin_skills/local/ by @yanyihan-xiaomi in #2297 Refactor - skill: move drive-mimo to repository scope (.mimocode/skills/) instead of the product-wide builtin bundle by @yanyihan-xiaomi in #2210 - session: simplify harness selection — explicit harness flag is the single source of truth; PTC transport no longer auto-selects the GPT toolset; MIMOCODE_CODEX_MODE=false forces the default harness for every model by @MiMoHardFather in #2243, #2252, #2303 - tool: remove standalone change_directory; require absolute working directories for shell/code-mode commands by @yanyihan-xiaomi in #2239 - prompt: rewrite Tools section as “Using your tools” guidance; instruction files default-on and decoupled from the dynamic system prompt flag by @MiMoHardFather in #2302, #2285 Internal / CI - session: retune the auto-overflow fixture for the flat 90% trigger by @yanyihan-xiaomi in #2266 - sdk: regenerate openapi and js sdk types (projection schema, harness description, compaction field docs) by @yanyihan-xiaomi in #2310 - chore: bump version to 0.1.14 by @yanyihan-xiaomi in #2311 Full Changelog: https://github.com/XiaomiMiMo/MiMo-Code/compare/v0.1.13...v0.1.14

WebCodr.jwtd

27 Aug 2026

5.2.05.6.0

VirusTotal:

## Faster rendering, identical output This release is internal work again. Nothing about jwtd's output changes, and on an ordinary token nothing about the speed is perceptible：a `jwtd <token>` run takes about 0.88 ms, and roughly three quarters of that is process startup and Go runtime initialization that happens before jwtd's own code runs at all. What changed is the printer. jwtd rendered its colored JSON with `go-prettyjson`, which marshalled the value back to JSON and re-decoded it before formatting, then assembled the result one string per token — each through `fmt.Sprintf`, each re-deriving its ANSI escape sequences. jwtd already holds the decoded token, so the round trip was pure overhead. The replacement walks the value straight into a single buffer, resolves each color's escapes once per section, and quotes strings through `encoding/json` only when quoting would actually change them. Where it is measurable — a 725 KB token, far larger than anything you would paste by hand: | | 5.5.0 | 5.6.0 | |---|---|---| | rendering the payload | 7.32 ms | 0.33 ms | | allocations while rendering | 19.6 MB, 136k | 579 B, ~0 | | decode and print, end to end | 11.4 ms | 4.3 ms | | decode and print, ordinary token | 14.6 µs | 4.0 µs | The last row is the work jwtd does for a normal token, and it is 1 % of the run. Startup dominates, as it did before. ## The output is byte-for-byte identical, and that is tested rather than asserted `go-prettyjson` is still in `go.mod`, now as a **test-only dependency**, so the formatter that replaced it can be held to its output byte for byte：over enumerated value shapes and 200 seeded-random JSON values, with color on and off. A second differential test pins the cached ANSI escapes against `fatih/color` itself — and earned its keep immediately, catching that the library's `Sprint` family closes with a reset per attribute while its `Fprintf` family closes with a plain reset. Conflating the two would have quietly changed the dimmed reason line under a failed verdict and the JWE encrypted-part listing. Beyond the test suite, 129 invocations of the 5.5.0 and 5.6.0 binaries across 15 tokens and 9 flag combinations produce identical bytes, exit codes and stderr included. ## Also in this release The escaping pass that sanitizes formatted JSON for the terminal had a fast path that never ran when color was on：it required printable ASCII, and colored output always carries the ESC bytes of its own color codes. It now admits them — everything that pass rewrites is DEL or above — which takes it from 612 MB/s to 4,214 MB/s on colored output, the interactive default. The repository also gains a benchmark suite covering the parse, render, escape, and signature paths, so performance claims about jwtd can be checked instead of argued. One of them exists to record a limit rather than to invite tuning：decoding the payload is now the largest remaining cost, and it is already at the standard library's floor for a decoder that must preserve exact numeric claim values. **Full Changelog**：https://github.com/webcodr/jwtd/compare/v5.5.0...v5.6.0