{"slug": "give-agents-homeplace-via-ssh", "title": "Give Agents Homeplace via SSH", "summary": "Agent Shell launches a hardened Linux box accessible via SSH or browser, giving AI agents like Claude, Codex, and Gemini real root access with gVisor sandboxing to prevent host compromise. The product targets developers who need persistent, monitored environments for AI agents to install packages, run services, and compile code, with live monitoring and preinstalled AI tooling.", "body_md": "### Real root, real apt\n\nYou are root. apt install anything, run services, compile whatever you want. A full Ubuntu box your agent controls, not a locked-down playground.\n\n**Agent Shell** is a hardened Linux box you reach over SSH\nor right in the browser. Give borg, Claude, Codex or Gemini real root,\nlet them apt install and build anything, and watch it live. Sandboxed so\nhard that even root inside cannot touch the host.\n\na real Agent Shell session · SSH or browser · watched live\n\nRuns the agents your team already trusts. Connects the platforms they work in. Built on infrastructure operated since 2009.\n\nHalf a minute on why a box that never turns off beats a sandbox that vanishes. The terminal you see is a real session, not a mockup.\n\nOne account and one AI budget for the whole team. Hand your agents a cloud VPS of their own, put the AI in your channels, or point it at your repo. From a free box to power shells, built for developers and the teams they ship with.\n\nA hardened Linux box you hand your agent root on, over SSH or the browser. gVisor-sandboxed so even root inside cannot touch the host, monitored live, with every AI agent and a QA toolkit preinstalled.\n\nLearn more →The always-on AI bouncer. Holds your connection around the clock and puts polish, summaries and agents inside IRC, Discord, Telegram, Slack and web chat.\n\nLearn more →The same AI as a coding agent in your shell. Hand it a task in plain language and it reads your repo, edits across files, runs your tests and checks its own work.\n\nLearn more →Three things a raw VPS will never give you. Real control, real containment, and eyes on everything your agent does.\n\nYou are root. apt install anything, run services, compile whatever you want. A full Ubuntu box your agent controls, not a locked-down playground.\n\nEvery box runs under gVisor, a userspace kernel between the box and the host, so even root inside cannot reach the host kernel. The isolation class Google Cloud Run and Replit trust for untrusted code.\n\nReal-time graphs for memory, CPU, disk and network, a running-agents indicator and a top-processes view. See exactly what your agent is doing, the moment it does it.\n\nSix layers of containment, live monitoring and every agent preinstalled.\n[See what's inside Agent Shell →](/agent-shell)\n\n```\npip install xshellz\npython\nfrom xshellz import Sandbox\nwith Sandbox.create() as sbx:\n    print(sbx.run(\"echo hi\").stdout)\npython\nfrom xshellz import Sandbox\nbox = Sandbox.get_or_create(\"my-agent\")   # same box every run\nbox.run(\"claude -p 'fix the failing test'\")\nnpm install xshellz\njs\nimport { Sandbox } from \"xshellz\";\nconst sbx = await Sandbox.create();\nconsole.log((await sbx.run(\"echo hi\")).stdout);\njs\nimport { Sandbox } from \"xshellz\";\nconst box = await Sandbox.getOrCreate(\"my-agent\");\nawait box.run(\"claude -p 'fix the failing test'\");\ngo get github.com/xshellz/xshellz-go\nsbx, _ := xshellz.Create(ctx, nil)\nr, _ := sbx.Run(ctx, \"echo hi\", nil)\nfmt.Println(r.Stdout)\nbox, _ := xshellz.GetOrCreate(ctx, \"my-agent\", nil)\nr, _ := box.Run(ctx, \"claude -p 'fix the failing test'\", nil)\nfmt.Println(r.Stdout)\nimplementation(\"com.xshellz:xshellz:0.2.0\")\ntry (Sandbox sbx = Sandbox.create()) {\n    System.out.println(sbx.run(\"echo hi\").stdout());\n}\njs\nSandbox box = Sandbox.getOrCreate(\"my-agent\");\nvar r = box.run(\"claude -p 'fix the failing test'\");\nSystem.out.println(r.stdout());\ngem install xshellz\nXshellz::Sandbox.create do |sbx|\n  puts sbx.run(\"echo hi\").stdout\nend\nbox = Xshellz::Sandbox.get_or_create(\"my-agent\")\nr = box.run(\"claude -p 'fix the failing test'\")\nputs r.stdout\ncomposer require xshellz/xshellz\nphp\n$sbx = Xshellz\\Sandbox::create();\necho $sbx->run('echo hi')->stdout;\n$sbx->kill();\nphp\n$box = Xshellz\\Sandbox::getOrCreate('my-agent');\n$r = $box->run(\"claude -p 'fix the failing test'\");\necho $r->stdout;\ncargo add xshellz\njs\nlet sbx = Sandbox::create(Default::default()).await?;\nlet r = sbx.run(\"echo hi\", Default::default()).await?;\nprintln!(\"{}\", r.stdout);\njs\nlet sbx = Sandbox::get_or_create(\"my-agent\", Default::default()).await?;\nlet r = sbx.run(\"claude -p 'fix the failing test'\", Default::default()).await?;\nprintln!(\"{}\", r.stdout);\n```\n\nSet `XSHELLZ_API_KEY`\n\nand you are live. Seven SDKs,\none behavior everywhere.\n[See what powers them, Agent Shell](/agent-shell).\n\nturborg holds your connection around the clock and adds an AI to your channels: polish a message before you send it, summarize what you missed, or hand a task to an agent. We have run IRC since 2009, and the same agent now works on Discord, Telegram, Slack and a hosted web chat.\n\nturborg · IRC connector · AI Polish · channel summaries\n\nFrom a modern web IRC client to AI automations you build by describing them - here's turborg, screen by screen.\n\nEvery screen is the real product. [Explore turborg →](/turborg)\n\n**borg** is the same AI pointed at your repo: a\nterminal-native coding agent that reads your codebase, writes the change\nacross as many files as it takes, runs your tests and checks its own\nwork. Run it in your own shell, or give it root inside an Agent Shell.\n\n`curl -fsSL https://turborg.com/install.sh | sh`\n\na real borg session · reads · edits · runs · verifies\n\nEvery plan includes an Agent Shell box, an always-on turborg with a daily AI budget, and access to borg. One price you can read top to bottom, no per-seat surprises.\n\n$ 0 /mo\n\n$ 10 /mo\n\n$ 24 /mo\n\n$ 60 /mo\n\nFull breakdown of boxes, tokens, limits and features on the [pricing page →](/pricing)\n\nxShellz is not a weekend project. The shells, runtimes and networks\nunder Agent Shell, turborg and borg have been online for over fifteen\nyears, and turborg is [open source](https://github.com/turborg/turborg). The intelligence is new. The ground it stands on is not.\n\nCan't find your answer? [Ask support →](/support)\n\nAgent Shell is a hardened Linux box you hand your AI agent root on, over SSH or right in the browser. It is a full Ubuntu machine where you can apt install anything, run services and build whatever you want, sandboxed with gVisor so even root inside cannot reach the host kernel. Live monitoring, egress and resource caps, and borg, Claude Code, Codex and Gemini plus a QA and eval toolkit all ship preinstalled.\n\nYes, that is the entire point. Every box runs under gVisor, a userspace kernel between the box and the host, so even root inside cannot reach the host kernel. Container root also maps to an unprivileged host user, and hard caps bound memory, CPU, processes, disk and network. A raw VPS with root gives you none of this.\n\nturborg is an always-on AI agent for real-time chat. It holds your connection around the clock so you never miss a message, and puts an AI in your channels: message polish before you send, one-click summaries when you have been away, and agents on command. IRC is our flagship connector, and turborg also runs on Discord, Telegram, Slack and a hosted web chat.\n\nborg is the same AI as a coding agent in your terminal. You give it a task in plain language and it reads your codebase, makes the change across as many files as needed, runs your tests and checks its own work. It runs in your own shell or inside an Agent Shell box. One account covers turborg, borg and Agent Shell.\n\nOpen-weight models (built on DeepSeek and Gemma) that we fine-tune, quantize and wrap in an agent harness ourselves. We do not pretrain from scratch and we are honest about that. The goal is AI that is fast, efficient and cheap enough to include in every plan.\n\nNo. xShellz runs the always-on, managed infrastructure we have operated since 2009, so there is nothing to deploy or babysit. Agent Shell boxes and turborg both run on it. If you would rather self-host turborg, the core is open source on GitHub.\n\nYes. The free plan needs no card and includes a free Agent Shell box, a real AI budget for turborg, and access to borg. When you outgrow it, Starter, Pro and Max raise the limits and keep everything always-on.\n\nNo card, no seat fee. One account gets you an Agent Shell box, an always-on turborg and borg in your terminal, live in minutes.\n\nChristina\n\nxShellz concierge", "url": "https://wpnews.pro/news/give-agents-homeplace-via-ssh", "canonical_source": "https://www.xshellz.com/", "published_at": "2026-07-18 19:56:19+00:00", "updated_at": "2026-07-18 20:21:12.645857+00:00", "lang": "en", "topics": ["ai-infrastructure", "ai-tools", "developer-tools", "ai-agents"], "entities": ["Agent Shell", "Claude", "Codex", "Gemini", "gVisor", "Google Cloud Run", "Replit"], "alternates": {"html": "https://wpnews.pro/news/give-agents-homeplace-via-ssh", "markdown": "https://wpnews.pro/news/give-agents-homeplace-via-ssh.md", "text": "https://wpnews.pro/news/give-agents-homeplace-via-ssh.txt", "jsonld": "https://wpnews.pro/news/give-agents-homeplace-via-ssh.jsonld"}}