monarchmaisuriya

Kernl — AI skill for Claude Code

AI community

A VS Code desktop extension: a coding agent (Deep Agents) that uses your own model key and works in VS Code's integrated terminal and integrated browser.

How to install Kernl

This entry records only its repository, not the path inside it, so there is no exact command to give. Open monarchmaisuriya/kernl and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Kernl does

A VS Code desktop extension: a coding agent (Deep Agents) that uses your own model key and works in VS Code's integrated terminal and integrated browser. It can also drive your installed Claude Code, Codex, or OpenCode. Every change it can make waits for your click, unless you pick a permission mode that allows it.

Alternatives in AI

  • Nodeterm — Node-based terminal manager for AI coding agents — tmux-backed terminals and parallel agent sessions as dragga 1.3k ★
  • Dinotty — Multi-device terminal server for AI coding agents 643 ★
  • Pi Agent Desktop — Pi — A cross-platform AI coding agent, bringing the Claude Code experience to your desktop 289 ★

README

kernl icon

kernl

A VS Code desktop extension: a coding agent (Deep Agents) that uses **your own model key** and works in **VS Code's integrated terminal and integrated browser**. It can also drive your installed Claude Code, Codex, or OpenCode. Every change it can make waits for your click, unless you pick a permission mode that allows it.

Install (team, from VSIX)

Download `kernl-X.Y.Z.vsix` from the [latest release](https://github.com/pushowl/kernl/releases/latest), then:

code --install-extension kernl-X.Y.Z.vsix --force

Or build it yourself:

npm ci && npm run package
code --install-extension kernl.vsix

What changed in each version: [CHANGELOG.md](CHANGELOG.md).

Requires VS Code 1.140 or newer. The integrated-browser tools are verified on VS Code 1.142.

First run

Open the **kernl** icon in the activity bar (or **Help > Welcome > Get started with kernl**). Pick an agent with the agent chip in the message box:

  • kernl: its own agent, with your model keys. Manage providers in the panel's ... menu stores a key and tests it; the model chip picks the provider and model.
  • Claude Code, Codex, OpenCode: your installed CLIs. Agent logins in the ... menu shows whether each is ready; a ready agent can be used in the shown workspace right away, and any agent can log in in a terminal or use a stored key.

For Ollama or another OpenAI-compatible server, use provider `openai-compatible` and `kernl.baseUrl` (e.g. `http://localhost:11434/v1`).

What kernl never does

  • Edit, delete, run a command, use the browser, or call an MCP tool without asking, unless your approval rules or permission mode allow it. See approvals and permission modes.
  • Store keys in settings: they live only in VS Code SecretStorage and are scrubbed from logs and panel messages.
  • Send traces: LangSmith tracing is forced off for every ru