Skip to content
Open
Show file tree
Hide file tree
Changes from 2 commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions docs.json
Original file line number Diff line number Diff line change
Expand Up @@ -194,6 +194,7 @@
"integrations/computer-use/yutori"
]
},
"integrations/docker-sandboxes",
"integrations/laminar",
"integrations/magnitude",
"integrations/notte",
Expand Down
44 changes: 44 additions & 0 deletions integrations/docker-sandboxes.mdx
Original file line number Diff line number Diff line change
@@ -0,0 +1,44 @@
---
title: "Docker Sandboxes"
description: "Run agents inside Docker Sandboxes with Kernel tooling and proxy-managed API auth"
---

The [Kernel kit](https://github.com/kernel/docker-sbx-kit) is a [Docker Sandboxes](https://docs.docker.com/ai/sandboxes/) [mixin](https://docs.docker.com/ai/sandboxes/customize/kits/) that gives any `sbx` agent:

- **Kernel CLI** (`@onkernel/cli`) installed at sandbox creation
- **Kernel agent skills** from [`kernel/skills`](https://github.com/kernel/skills), so Claude Code (and any agent that reads `~/.agents/skills`) can drive Kernel without prompting
Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Specific third-party AI product names used in docs

Medium Severity

The new page references "Claude Code" on line 9 and "Claude" on lines 26 and 42. Per project rules, documentation must not reference specific third-party AI products by name — generic, vendor-neutral terms like "a coding agent," "the agent," or "your AI tool" are expected instead. "Claude Code" is explicitly called out as an example of a term to avoid.

Additional Locations (2)
Fix in Cursor Fix in Web

Triggered by learned rule: Use generic terms instead of specific third-party AI product names

Reviewed by Cursor Bugbot for commit 42443ca. Configure here.

- **Proxy-managed `KERNEL_API_KEY`** — your real key stays on the host. The `sbx` proxy injects it as `Authorization: Bearer …` on requests to `api.onkernel.com`. The agent inside the sandbox never sees the secret.

The last point is what makes this integration worth using over `npm install -g @onkernel/cli` inside a custom kit.

## Quickstart

```bash
export KERNEL_API_KEY=sk-kernel-...
export ANTHROPIC_API_KEY=sk-ant-...

sbx run --name kernel-demo \
--kit "git+https://github.com/kernel/docker-sbx-kit.git" \
claude -- \
"Using the Kernel CLI, create a browser and navigate to news.ycombinator.com. Tell me the top five articles."
```

Claude calls `kernel` inside the sandbox → CLI hits `api.onkernel.com` → the `sbx` proxy attaches your `KERNEL_API_KEY` → the request authenticates as you.

The kit's full `spec.yaml`, install commands, and allowed domains live in the [repo README](https://github.com/kernel/docker-sbx-kit#readme).

## Prerequisites

- `sbx` installed and signed in — see [Docker's getting started guide](https://docs.docker.com/ai/sandboxes/get-started/)
- A Kernel API key from the [Kernel Dashboard](https://dashboard.onkernel.com/api-keys)
- An Anthropic API key from the [Anthropic Console](https://console.anthropic.com/) if you're using the built-in `claude` agent

## Customizing or extending

For everything not specific to Kernel — loading kits from local paths or OCI registries, stacking multiple mixins, building your own agent kit, debugging the proxy, `sbx kit add` for running sandboxes — see [Docker's kit reference](https://docs.docker.com/ai/sandboxes/customize/kits/). The Kernel kit is a standard mixin and composes with anything else you put on top.

## Next steps

- Browse [`kernel/skills`](https://github.com/kernel/skills) to see what Claude can do out of the box
- Read the [Kernel CLI reference](/info/cli) for commands available inside the sandbox
- Learn about [browser creation](/browsers/create-a-browser), [stealth mode](/browsers/bot-detection/stealth), and [Profiles](/auth/profiles) for harder automation targets
1 change: 1 addition & 0 deletions integrations/overview.mdx
Original file line number Diff line number Diff line change
Expand Up @@ -30,6 +30,7 @@ Kernel provides detailed guides for popular agent frameworks:
- **[Agent Browser](/integrations/agent-browser)** - Browser automation CLI for AI agents
- **[Browser Use](/integrations/browser-use)** - AI browser agent framework
- **[Claude Agent SDK](/integrations/claude-agent-sdk)** - Run Claude Agent SDK automations in cloud browsers
- **[Docker Sandboxes](/integrations/docker-sandboxes)** - Add Kernel tooling and proxy-managed auth to Docker sandboxes via the Kernel kit
- **[Stagehand](/integrations/stagehand)** - AI browser automation with natural language
- **[Computer Use (Anthropic)](/integrations/computer-use/anthropic)** - Claude's computer use capability
- **[Computer Use (OpenAI)](/integrations/computer-use/openai)** - OpenAI's computer use capability
Expand Down
Loading