Skip to content

docs: document Claude Code local model workflows #25641

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
dvdksn wants to merge 2 commits into
base: main
Choose a base branch
from
+44 −7
Open

docs: document Claude Code local model workflows #25641

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
0db5506
docs: document local model workflows
dvdksn Jul 24, 2026
4da1a47
docs: clarify local model platform support
dvdksn Aug 5, 2026
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
51 changes: 44 additions & 7 deletions content/manuals/ai/sandboxes/agents/claude-code.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,9 +2,9 @@
title: Claude Code
weight: 10
description: |
Use Claude Code in Docker Sandboxes with authentication, configuration, and
YOLO mode for AI-assisted development.
keywords: docker sandboxes, claude code, anthropic, ai agent, sbx
Use Claude Code in Docker Sandboxes with authentication, local models,
configuration, and YOLO mode for AI-assisted development.
keywords: docker sandboxes, claude code, anthropic, ai agent, sbx, local models, llmman, ollama
---

Official documentation: [Claude Code](https://code.claude.com/docs)
Expand Down Expand Up @@ -112,8 +112,45 @@ this base.

## Use a local model

To run Claude Code in a sandbox against a local model on your host through
Docker Model Runner, see
The `--model` flag routes Claude Code's Anthropic API requests to a model
served on your host. This feature is experimental and isn't supported on
Windows.

@ericcurtin ericcurtin Aug 5, 2026 •
edited
Loading

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

Could we change this to only "This feature is experimental."? I don't want to set the expectation that Windows is out because it's not, it's an important platform. Could be a follow on PR.

@dvdksn dvdksn Aug 5, 2026

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

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

Hmm, I heard that we are not shipping llmman + llama in Windows MSI. I took that as Windows is not supported. Can you confirm? cc @ilopezluna

@ilopezluna ilopezluna Aug 5, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

it's not, i can confirm 👍

@ericcurtin ericcurtin Aug 5, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

@ilopezluna it's not supported for just this release?

It should at least say windows support is pending, not that we are not supporting it.

@ilopezluna ilopezluna Aug 5, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

I can confirm its not supported in this release we might say something like "coming soon" that's what we did in DMR. So being clear that is not supported I think its fine and also mentioning that will be here eventually so we set the right expectations

@ericcurtin ericcurtin Aug 5, 2026 •
edited
Loading

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

Local models IMO isn't supported at all, experimental means, have fun, but don't expect any support yet, but approved regardless, we might be too early even documenting this. But it's done now, we should merge and avoid wasted effort.


Enable the feature:

```console
$ sbx settings set platform.allowExperimentalFeatures true
$ sbx settings set feature.model true
```

To use the bundled `llmman` model server, pass a GGUF model reference or short
name:

```console
$ sbx run --model gemma4 claude
```

On first use, `sbx` starts `llmman`, pulls the model, and leaves the server
running on your host. Later sandboxes reuse the server and its model store.

To use an existing Ollama installation instead, prefix the model name with
`ollama/`:

```console
$ sbx run --model ollama/gemma4 claude
```

Ollama must already be installed and running. `sbx` connects to it but doesn't
start or manage the Ollama process.

You can also change the model for an existing sandbox:

```console
$ sbx run --name <sandbox-name> --model <model-name>
```

Changing the model recreates the sandbox container. The workspace and
kit-owned volumes persist.

To use Docker Model Runner instead, see
[Run Claude Code in a Docker Sandbox with Docker Model Runner](/guides/claude-code-sandbox-model-runner/).
For the host-only version without a sandbox, see
Comment thread
dvdksn marked this conversation as resolved.
[Use Claude Code with Docker Model Runner](/guides/claude-code-model-runner/).