An open API service indexing awesome lists of open source software.

https://github.com/maresb/claude-code-pixi-cloud-environment


https://github.com/maresb/claude-code-pixi-cloud-environment

Last synced: 6 days ago
JSON representation

Awesome Lists containing this project

README

          

# claude-code-pixi-cloud-environment

Transparently activate a [pixi](https://pixi.sh) environment for **every command
Claude runs** in a [Claude Code on the web](https://code.claude.com/docs/en/claude-code-on-the-web)
cloud session — so `python`, `pip`, `pytest`, and friends resolve to the project's
**fully activated** pixi env (`CONDA_PREFIX`, `XML_CATALOG_FILES`, …), with no
`pixi run` prefix and nothing repo-specific to configure.

## Setup

Open your environment's settings in the Claude Code web UI and configure two things.

![The Update cloud environment settings dialog in the Claude Code web UI](./cloud-environment-settings.png)

**1. Setup script** — paste the contents of
[`cloud-environment-setup.sh`](https://raw.githubusercontent.com/maresb/claude-code-pixi-cloud-environment/main/cloud-environment-setup.sh).

**2. Network access** — `pixi install` must reach the installer and conda channels.
Choose **Custom**, keep **"Also include default list of common package managers"**
checked (covers PyPI, etc.), and add:

```
*.anaconda.org
*.pixi.sh
*.prefix.dev
anaconda.org
pixi.sh
prefix.dev
release-assets.githubusercontent.com
```

No environment variables are required — the wrapper does all activation. (Do *not*
rely on `BASH_ENV`; the harness strips it.)

## How it works

The setup script installs pixi, builds the project's environment, and wraps
`/bin/bash` so Claude's shell commands enter that environment automatically.

See [`DESIGN.md`](./DESIGN.md) for the full constraints, the approaches that failed,
the logging, verification checks, and limitations.

> ⚠️ **Largely LLM-generated, and it replaces the system `/bin/bash`.**
> `cloud-environment-setup.sh`, this README, and [`DESIGN.md`](./DESIGN.md) were
> produced mostly by an LLM (Claude) through iterative testing in a cloud container.