Chat with your CloudKita agent right inside VS Code — same local tool access
(read/write files, run terminal commands, project graph search) as the kita
CLI, with live visibility into what the agent is doing (reasoning, tool calls,
results) as it works.
Features
- Chat with any of your CloudKita agents directly from the editor sidebar.
- Local tool execution ("remote brain, local hands") — the agent can read/ write files and run terminal commands on your machine, with your explicit approval for every write/edit/command.
- Live activity view — see the agent's reasoning, tool calls, and results stream in as they happen, not just the final answer.
- Project Graph — a local, persistent index of your workspace's symbols
and imports, visualized as an interactive graph (
CloudKita: Show Project Graph), so the agent (and you) can understand project structure without repeated file scans. - Permission modes —
prompt(ask before every write/exec, recommended), oryolo(auto-approve, for sandboxes only). Dangerous commands always require confirmation regardless of mode.
Getting Started
- Install the extension.
- Open the CloudKita icon in the Activity Bar and sign in (browser login or a workspace API key from your CloudKita dashboard).
- Select an agent and start chatting.
Requirements
- A CloudKita account with at least one deployed agent.
Settings
| Setting | Description |
|---|---|
cloudkita.apiBase |
Backend API base URL (only change for self-hosted/staging). |
cloudkita.webBase |
Dashboard base URL, used for browser sign-in. |
cloudkita.localTools.mode |
Permission mode for local file/terminal access: prompt (default), auto-read, or yolo. |
cloudkita.debug |
Log connection/handshake diagnostics to the CloudKita output channel. |
Privacy & Security
Local tool calls (file writes, edits, and terminal commands) always require your explicit approval by default. Path access is scoped to your open workspace folder. See cloudkita.id for our full privacy policy and terms of service.