Claude Code where the lid never closes.

Claude Code is the strongest coding agent shipping in May 2026: Claude Opus 4.7 leads Terminal-Bench 2.0 around 69% and posts SWE-bench Verified in the high 80s, paired with the deepest MCP integration in the industry and a per-action permissions model that's actually thoughtful about gating. Anthropic's "Remote Control" feature (released February 2026) lets you continue a local Claude Code session from a phone, tablet, or browser. There's only one problem with running Claude Code on your laptop: your laptop closes.

Move the session itself onto a server that doesn't.

What changes when Claude Code lives on a workstation#

Three things stop being painful:

1. Eight-hour runs become normal. Kick off the big refactor at 11pm. The agent works through the night. You wake up to a draft PR.
2. Multiple sessions at once. Two Claude Code sessions on different branches. No port collisions, no state collisions, no "wait, which window was the deploy?"
3. Passkey-gated production access. The agent never sees your ~/.aws or GITHUB_TOKEN. They live in a server-side vault. When a deploy needs them, the action pauses, you tap your passkey, the credential is brokered through a separate process.

How to set it up#

The setup is five minutes if you already have a Claude subscription.

1. Sign up. Hobby ($20/mo) or Pro ($50/mo). No credit card for the trial.
2. Bring your Claude key. Paste your Anthropic API key into Ellul's vault, or your Claude Pro/Max OAuth credential if you're on the subscription path. The key never leaves the vault.
3. Start a workstation. A persistent sandbox provisions in roughly 90 seconds, pre-installed with Claude Code, Codex CLI, OpenCode, Cursor's CLI, and Grok Build.
4. Connect a passkey. FIDO2 from your phone, laptop, or hardware key. This is what gates git push, vercel deploy, and any other privileged action.
5. Open a session. Either through Ellul's chat UI in your browser, or by SSHing into the workstation and running claude directly.

That's it. The session is yours until you cancel.

Pricing: flat, not metered#

What it isn't#

When you'd use it#

The shape of work that changes most when you move Claude Code onto a workstation:

• Long unattended refactors. Migrations, dependency upgrades, codebase-wide renames. The agent works while you sleep.
• Production-touching agents. Anything where the agent runs git push, opens PRs against main, or talks to a real database.
• Multiple parallel sessions. Coding, reviewing, documenting: three agents in three sandboxes, with read-only peering so the reviewer can see the coder's source without being able to write to it.
• Phone-driven approvals. You're on the subway. Push notification. Tap your passkey. The agent's PR is merged.

See also#

• Claude Code vs Cursor →
• Claude Code vs Ellul →
• Why we built the gate outside the agent process →

Does Ellul include a Claude subscription?

No. Ellul is BYOK (bring your own key). Use your existing Claude Pro or Claude Max subscription. Ellul charges only for the workstation: $20/mo Hobby or $50/mo Pro.

Will Claude Code keep working when I close my laptop?

Yes. The agent runs on a persistent workstation; your laptop is just a thin client. Close it, walk away, come back in eight hours and read the agent's output.

Can the agent read my Anthropic API key?

No. The key sits in a server-side vault and is brokered into the agent process for the duration of a single tool call. The agent process never sees the key file directly.

What about MCP servers?

Ellul runs MCP servers either inside the workstation (the way you'd run them locally) or as long-lived server-side processes outside the agent's namespace. Both are gated.

How is this different from Anthropic's Remote Control feature?

Remote Control sends messages from another device into a Claude Code session that still lives on your laptop. The session itself stops when the laptop sleeps. Ellul moves the session onto a server that doesn't sleep.