1. Conduid
  2. Developer Tools
  3. Sandboxed.sh
MCP server · Developer Tools

Sandboxed.sh

All your AI providers, machines, and agent harnesses in one place. Formerly sandboxed.sh (now the backend to the Orb clients).

72Good

Scored 8 hours ago · breakdown

About Sandboxed.sh

Sandboxed.sh is an MCP server published by Th0rgal in the Developer Tools category: all your AI providers, machines, and agent harnesses in one place. Formerly sandboxed.sh (now the backend to the Orb clients).

The repository has 530 stars and 57 forks, with the last commit 15 hours ago. Recent commit activity is the strongest signal that reported issues will get attention.

Install

From source
git clone https://github.com/Th0rgal/orb

This server has no ConduID identity, so agent calls to it are not receipted. Pin the version you install and review the source before granting it credentials.

Ask AI

✦

Ask AI about Sandboxed.sh

Powered by Claude · Grounded in docs

I know everything about Sandboxed.sh. Ask me about installation, configuration, usage, or troubleshooting.

Security checks

Static checks derived from the repository 8 hours ago. Heuristics on the source, not an audit.

  • ✓README presentThe repository ships documentation.
  • !License declaredNo license file. Legally you have no right to use it in a product.
  • ✓Tests presentA test suite exists in the repository.
  • ✓Dependencies pinnedA lockfile pins dependency versions.
  • !Scoped permissionsDoesn't declare a permission scope. Assume it can do anything its process can.

1 check has not been run yet.

Releases

v0.12.0v0.12.0 — Grok Build, Needs You, iOS UX · 16 May 2026v0.12.0 — Grok Build, Needs You, transcript streaming, iOS UX Grok Build — first-class backend Full xAI OAuth flow (device-auth), `~/.grok/auth.json` mirrored per mission workspace, container-aware `--cwd` translation, `grok-build` default…
v0.11.5v0.11.5 - Native Harness Loops & iOS Network Resilience · 14 May 2026Bundles PR #432 — clean reimplementation of the still-unique work from the abandoned #430 (Codex goal mode + iOS network resilience WIP). Features Native harness loops as Automation rows Claudecode and codex `/goal` continuation loops now…
v0.11.1v0.11.1 - Mission Workbench BugBot Fixes · 3 May 2026v0.11.1 highlights Patch release for the Mission Workbench and Needs You inbox rollout. Changes Aligns Mission Workbench running-state rendering with the header's effective running-state logic instead of treating any `RunningMissionInfo`…
v0.11.0v0.11.0 - Mission Workbench & Goal Mode · 3 May 2026v0.11.0 highlights (since v0.10.0) This release brings the dashboard mission-management workflow forward and includes the latest backend/runtime reliability work. Most important changes Added the Mission Workbench side panel on the control…
v0.10.0v0.10.0 · 23 Feb 2026v0.10.0 highlights (since v0.9.1) This release focused on mission reliability, automation control, and cross-client parity. Most important changes Hardened OpenCode/Codex execution paths: workspace-specific CLI/binary resolution, outdated…

README

Orb on macOS, following a Verity coding mission

Choose where your agent works

Mode Where it runs How you use it
Your computer A local coding agent on your desktop Choose New Agent → This computer, then a harness and model. Work with your local files and tools.
Your private cloud Your own servers and remote machines Add machines to sandboxed.sh, then select one in Orb. Run agents in the configured host or isolated container workspace.
Cloud agents A coordinator or provider-managed assistant Choose Cloud agent, then Hermes, ChatGPT, Grok Bot or Cursor Cloud. Continue durable or provider conversations from Orb.

Use Claude Code, Codex, Antigravity, OpenCode or Grok for local and remote work, with live model, effort and cyber-access controls where supported. Cloud agents have their own connections: Hermes binds to durable coordinator conversations and router chains; ChatGPT uses a signed-in browser profile and your available subscription modes; Grok Bot uses its connected account; Cursor Cloud uses its official API with a repository, Git reference and model.

Creating an agent in Orb on macOS

Pick up the conversation anywhere

Projects organize missions, per-project colors, skills and a synced local Context folder in Finder. Inside a conversation:

  • Follow and steer live runs: collapsible Worked — … tool/thought folds, nested subagents and callback reviews, inline prompt editing, and a Cursor-style follow-up queue that sends on turn completion or immediately.
  • Ask on the side: open the docked /btw side-agent panel (⌘⇧J) to ask questions with shared workspace context without interrupting the main turn.
  • Keep long work moving: schedule durable mission wake-ups, automatically resume interrupted runs after restarts, and transfer workspaces across machines (preserving symlinks while skipping build artifacts).
  • Rich output: Markdown, code, tables, LaTeX, quizzes, inline images and clickable file links stay readable inside the transcript.

Orb for iOS connects to the same sandboxed.sh server: browse projects with synced colors, inspect mission transcripts and activity folds, send follow-ups, view or edit shared Markdown context, and manage backend, machines and providers.

Bring your subscriptions

Sign in to supported subscription accounts directly from Providers in Orb (macOS, web and iOS) or connect them through CLIProxyAPI Plus. You can authenticate via OAuth, renew tokens, enable or disable individual accounts, and inspect live plan/quota status; the proxy handles routing and rotation across eligible accounts while Orb lets you choose the harness and model.

This routes model inference for coding agents. Managed cloud agents use separate service adapters; connecting a proxy account does not sign in to the ChatGPT browser or create a Cursor Cloud API account.

Get started

  1. Run sandboxed.sh. Follow the Docker guide or native Linux guide. It keeps the project record, mission history and remote execution services.
  2. Open Orb. Build the desktop client or the iOS app, then enter your server URL and sign in.
  3. Connect your execution environment. Set up local harnesses, register remote machines, or connect a cloud-agent account. Configure inference providers for the models you want to use.
  4. Create a project and launch an agent. Choose its execution mode, write your task, and keep its conversation and results together.

Under the hood

Orb is the client; sandboxed.sh (the project's former name) is now the backend to the Orb clients. Desktop is built with Tauri and SolidJS; iOS uses SwiftUI. The Rust backend owns projects, missions, event history and remote-node coordination. Cloud adapters observe provider work independently of the client window.

The same control plane is available over MCP for coordinators such as Hermes. Automation uses the same project and mission records as Orb.

Documentation index · Execution architecture · Workspaces · MCP and Hermes · Development and troubleshooting


Orb (orb.thomas.md) · Formerly sandboxed.sh (now the backend to the Orb clients).

README fetched from the repository 8 hours ago. The original is authoritative.

Questions

About Sandboxed.sh

How do I install Sandboxed.sh?

No package has been published for this server, so install it from source: clone the repository and follow its README.

Is Sandboxed.sh safe to use with an AI agent?

Its trust score is 72 out of 100 (good). It passes 3 of the 5 static security checks run so far; the failures are listed above. It has no ConduID identity yet, so agent calls to it are not receipted.

Is Sandboxed.sh still maintained?

Yes — the latest release is v0.12.0 (16 May 2026), and the last commit was 15 hours ago. The repository has 530 stars and 18 open issues.