Question 1

What is CloudAgent?

Accepted Answer

CloudAgent is an always-on cloud computer for terminal-based AI coding agents. You SSH in from any terminal, start agent tasks, disconnect whenever you want, and come back to logs, diffs, and PR-ready work.

Question 2

Is this an IDE?

Accepted Answer

No. There is no proprietary editor or required extension. Your terminal is the interface; IDE integrations may come later.

Question 3

Do I need to use a special terminal?

Accepted Answer

No. Anything that can SSH works: Terminal.app, iTerm2, Ghostty, Warp, Windows Terminal, the VS Code terminal, or an SSH app on your phone or iPad.

Question 4

What happens when I disconnect?

Accepted Answer

Your session and its tasks keep running. When you reconnect — from the same device or a different one — you reattach to the same session and review what happened while you were away.

Question 5

Which coding agents are supported?

Accepted Answer

The launch roster is Claude Code, OpenAI Codex CLI, Gemini CLI, Aider, and OpenCode, plus selected open-model tooling. Support levels are labeled honestly per agent on the Agents page.

Question 6

Can I use my own API keys?

Accepted Answer

Not during the beta. Managed credits are the default path; bring-your-own-key is planned for advanced users and enterprise after the beta.

Question 7

How do credits work?

Accepted Answer

1 credit = $1 of managed model balance. You buy prepaid packs, and model usage debits credits at published per-token rates. No subscription during the beta.

Question 8

What do credits pay for?

Accepted Answer

Managed model token usage — input, cached input, and output tokens at the rates on the pricing page.

Question 9

Are cloud workspaces included?

Accepted Answer

During the private beta, workspace compute is included under fair-use limits. Credits pay for model tokens only.

Question 10

Can I use this with private repos?

Accepted Answer

Yes — that's the primary use case. Repo access is scoped per workspace, and the security model is built around not giving agents more access than the task needs.

Question 11

Is the service secure enough for company code?

Accepted Answer

We're building around least privilege, task isolation, scoped secrets, and audit logs — and we're explicit about what we don't claim yet (no SOC 2 or HIPAA certification during beta). Read the security page for the honest picture, or email wu@learnwithpal.org for a security review call.

Question 12

How does invite-only access work?

Accepted Answer

You request an invite on the waitlist with a bit of context about how you use coding agents. We manually invite users in cohorts while the service scales; accounts are created by the team, never self-served.

Question 13

Can teams use it?

Accepted Answer

Yes — small team pilots are part of the beta. Team budgets and shared controls are on the roadmap; tell us about your team in the waitlist form.

Question 14

What happens if an agent burns too many tokens?

Accepted Answer

Tasks run with hard budgets and stop when they hit them. Low balances trigger warnings, and below 1 credit new managed-model tasks won't start.

Question 15

Can I get a refund?

Accepted Answer

Credits are refundable where required by law and otherwise at our discretion — email support and we'll look at it case by case. Promotional credits are never charged for, so there's nothing to refund.

Questions, answered plainly.

Request access to the private beta.