Kilo Code Cloud Agents: Remote Local Sessions and Multi-Agent Coordination
Published on 10.04.2026
Kilo Code Cloud Agents Get Remote Local Connection and Multi-Agent Coordination
TLDR: You can now connect Kilo Cloud Agents directly to a running local session via /remote, giving you a cloud window into local compute from anywhere. A new multi-agent coordination pattern lets you run parallel sessions through a single dashboard.
Summary: The core tension in AI coding agents has always been the cloud-vs-local split. Cloud agents give you accessibility from any device and sessions that don't stop when you close your laptop. Local agents give you full access to your codebase, your environment, and your compute. Kilo Code's latest update collapses that split in a specific way: instead of running your code in the cloud, your local machine handles the compute, and the cloud gives you a persistent window into it.
The implementation is straightforward in concept. You start a session in the Kilo CLI, run /remote to enable remote mode, and your active session appears in the Cloud Agents dashboard. From there, you can check on progress, answer questions the agent surfaces, and approve permission dialogues — all from a browser or phone. The two-way connection means this isn't read-only monitoring; it's full interaction with the same session. The permission model is preserved: when the agent wants to do something sensitive, the dialogue appears wherever you're connected, not just in the CLI.
The practical use case this enables is the "leave your desk mid-session" scenario that anyone running long agent tasks has encountered. You start something that will take 20 minutes, need to step away, and currently you're either waiting at your desk or accepting that the session might stall waiting for your input. Remote mode means you can answer the agent's questions from your phone and let it keep running.
What I find more interesting from an architectural standpoint is the multi-agent coordination pattern the update enables. The article describes using the dashboard as a "control panel" for running multiple agents in parallel — one agent adding missed translations to a PR, another doing a multi-perspective code review on a different PR, a third building something in the cloud while you test locally. The key insight is that cloud sessions don't stop when you close your laptop, so you can delegate "run this until it's done" tasks to cloud agents and focus locally on the work that needs your attention.
The quality-of-life improvements deserve mention too. The dashboard consolidation — putting session creation and session history on a single page rather than split across two — sounds minor but makes the product significantly easier to use as a daily driver. Sessions grouped by day and the ability to delete completed sessions are the kinds of details that matter when you're using a tool dozens of times a week.
Key takeaways:
- Remote mode connects cloud dashboard to a running local CLI session via
/remote - Two-way interaction: answer questions, approve permissions from browser or phone
- Multi-agent pattern: run parallel cloud sessions for different tasks while working locally on one
- Dashboard consolidated to single page with day-grouped history and session deletion
Why do I care: The remote local session feature is solving a specific real problem I run into constantly. Long-running agent tasks that require occasional input are the current worst-case scenario for AI coding workflows — you're either tied to your desk or you lose momentum when the agent stalls waiting for you. The /remote command addresses that directly. The multi-agent coordination pattern is the more ambitious vision: using cloud sessions as a delegation layer while you focus locally, which is how I want my AI tooling to work at scale.