Publish administrator termination to session readers and reject new input before rebuilding compute. Preserve normal recovery after sandbox reclamation, explicit manual recovery, and completed history.
Cover both lifecycle paths with live tests and give deployments a stable hook for styling the recovery control.
Co-Authored-By: Claude <noreply@anthropic.com>
Give visitors a direct trial from the homepage, navigation and both README introductions, with the signup and session limits visible before entry.
Co-Authored-By: Claude <noreply@anthropic.com>
Strengthen sandbox isolation and authentication, make all five engines work
through the bundled installer, and preserve conversations across sandbox and
service restarts. Add team login and single-container deployment, with upgrade
instructions for replacing existing 0.1.0 sandboxes.
Co-Authored-By: Claude <noreply@anthropic.com>
- Homepage, README and overview introduce the astrabox CLI: configuration
as code (export, diff, apply), tasks from the command line, JSON output
for scripts and coding agents, and the same operations as MCP tools.
- Homepage and README list the nine plugin interfaces, their entry-point
groups, built-in implementations and conformance suites.
- On phones the docs menu opened as a 59px strip: the navbar's
backdrop-filter made it the containing block of Docusaurus's fixed drawer
(facebook/docusaurus#6996). The blur now applies only at desktop widths,
and a phone-viewport test covers the menu and the page contents.
- Quoted shell strings in light code blocks meet AA contrast.
The open-source, self-hosted alternative to Claude Managed Agents: run
Claude Code, Codex, Hermes, DeepSeek Harness and Pi as managed Agents on
your own infrastructure, with conversations that start and resume in
seconds.
