In-depth analysis · About 5 min read
Kimi Code’s October update: Linux access and clearer session boundaries
Desktop 1.1.0 and CLI 2.2.0 add workflow controls while changing progress, interruption, hooks and forked-session behavior.
- Source published
- Published here
- Source checked
Kimi’s official changelogs date Desktop 1.1.0 and CLI 2.2.0 to October 14. We treat them as one coding-tool update, rather than two model launches. Linux packaging, remote control and better work visibility are the visible additions; message handling and session-boundary fixes matter just as much for existing workflows.
Listen to the news
Read this article aloud or keep listening to the news.
Preparing the reader…
Device voice
Speech is handled by your browser or device. AISubHub does not send the text to an audio API. Voice availability and playback with a locked screen vary by device.
Resuming may repeat a short phrase.
At a glance
- Desktop adds Linux AppImage, deb and RPM packages. Remote Control must be enabled in Settings; progress updates are off by default in both Desktop and CLI.
- In CLI Tower mode, a message received while the agent is running aborts the current turn and runs as its own turn. Forked sessions no longer fire cron tasks belonging to the source session.
One release event with a wider Linux entry point
Desktop 1.1.0 adds Linux installers in AppImage, deb and RPM formats. The same entry introduces a right-sidebar File Tree for browsing and previewing workspace files, and lists fixes for missing line-by-line diffs, stuttering on large files and a crash on some Macs when chat contains emoji. CLI 2.2.0 accompanies the desktop release on the same date. Our interpretation is that the meaningful change is broader access to a shared coding workflow, not a new underlying model. Package availability alone does not establish compatibility with every distribution or hardware setup.
Remote control still operates a local agent
The Desktop changelog says a link or QR code can let another device control the local agent, after enabling Remote Control in Settings. It also lists in-app browser download progress, agent-handled popups and pausable agent control. These describe ways to observe and intervene in work; they do not say that the local agent becomes a hosted cloud service. Our suggested operational boundary is to confirm which local workspace is active and keep appropriate permissions and review. That advice is not an independently tested access-control claim about this release.
Progress is optional, with different controls in each client
Desktop can display mid-task progress after enabling it in Settings → Lab; it is off by default. CLI adds the opt-in NotifyUser progress feature for new sessions through the TUI experiments panel, also off by default. Both add the kimi app command to open a new desktop chat in the current or specified directory. CLI additionally keeps a one-line user-message summary visible when the message scrolls away in fullscreen mode, with a jump back on click. Our reading is improved visibility and handoff between interfaces, not proof that the agent’s progress statements accurately establish completion.
Message arrival and interruption are not the same behavior
Desktop says queued messages, or messages sent while the agent waits on background tasks, now arrive immediately. The CLI Tower-mode entry describes a different rule: a message sent during execution aborts the in-progress turn and runs as a separate turn. These should not be flattened into a promise of uninterrupted parallel messaging. The CLI also shows a shutdown status while exit hooks run, with Ctrl+C forcing immediate exit instead of waiting for their timeout. Teams relying on in-flight work or cleanup hooks should test those transitions; this is our upgrade advice, not a reported test.
Hooks and invalid configuration need migration attention
CLI fixes misplaced turns, lost replies and rendering problems when a UserPromptSubmit hook injects text. Hook JSON lacking a message field is no longer added to the conversation. It also removes the never-implemented executor = kaos option from MCP stdio server configuration; files still using it now fail validation at load time. These are concrete behavioral boundaries, not a blanket statement that hooks or MCP have been removed. Our recommendation is to inspect existing configuration and representative hook outputs before upgrading, preserving intended message content instead of assuming every formerly accepted setting was functional.
Forked sessions should not duplicate scheduled work
CLI says cron tasks from the source session no longer fire inside a forked session. This is a fix to inheritance across a session boundary, not a claim that all scheduled tasks are disabled. A bounded acceptance check could fork a noncritical session, verify task ownership, send a message during a controlled run and inspect hook output and changes. The Desktop browser’s pause control and diff display can support review, but do not replace it. We did not perform these tests. Preserve normal rollback and confirmation practices rather than infer universal safety from a changelog.
Why it matters
The update is relevant to teams moving between terminal and desktop or reviewing long agent tasks. A useful upgrade check tests message timing, hooks, configuration and scheduled work, rather than just whether the application opens.
Limits and availability
We read both complete dated release entries, without installing or running the products. These notes do not announce a new model, new pricing or free usage, and do not amount to a security audit or a guarantee that every workflow is unaffected.
Sources and evidence
Source published: Oct 14, 2026 · Source checked: Oct 20, 2026
Source published: Oct 14, 2026 · Source checked: Oct 20, 2026
AI-assisted editorial articles checked against linked primary sources. Vendor claims are attributed; inclusion is not endorsement.