cloudagent

$ invite-only private beta

Your AI development computer in the cloud.

SSH from any terminal. Run coding agents in persistent sessions. Disconnect anytime. Come back to logs, diffs, and PR-ready work.

Request inviteSee pricing

Invite-only private beta. No public registration. Built for developers who live in the terminal.

any terminal — ssh
[dev-api] 1 task · attachedcloudagent
Illustrative beta session

the problem

Coding agents should not depend on your laptop staying awake.

Local sessions die

Sleep, network drops, battery, reboot, lost terminal history.

Agent setup is fragmented

Every CLI agent has its own installation, auth, config, and model setup.

Costs are opaque

Long-running agent tasks can burn tokens without clear task-level visibility.

Security is too broad

Agents often get more repo, shell, network, and secret access than they need.

how it works

Connect. Delegate. Disconnect. Review.

  1. 01 · connect

    Connect from any terminal

    SSH into your CloudAgent workspace.

  2. 02 · delegate

    Start an agent task

    Use Claude Code, Codex CLI, Gemini CLI, Aider, OpenCode, or the default task wrapper.

  3. 03 · disconnect

    Leave it running

    Sessions persist when you close your laptop or switch devices.

  4. 04 · review

    Review the result

    Inspect logs, token cost, changed files, tests, and PR-ready output.

  1. [queued]

    Task accepted

  2. [running]

    Agent working

  3. [needs input]

    Approval requested

  4. [detached]

    You disconnect — work continues

  5. [done]

    Diff, logs, PR ready

capabilities

Everything an AI coding agent needs, already there.

Any terminal

SSH-first access from the terminal you already use.

Persistent sessions

Long-running work survives disconnects.

Preinstalled agents

Major CLI coding agents maintained and ready.

Fast repo boot

Devcontainers, package caches, dotfiles, and common runtimes.

One credit balance

Token usage across supported models from one account.

Safe autonomy

Branch isolation, scoped secrets, approvals, and audit logs.

Task history

Transcripts, diffs, usage, and status preserved per task.

PR-oriented output

Agent work ends in reviewable diffs and pull requests.

Explore all features

bring your agent

Bring your favorite coding agent.

CloudAgent is designed for terminal-native coding agents, not a single proprietary workflow.

Claude Code

[planned]

Strong default for agentic coding workflows.

Codex CLI

[planned]

Important for OpenAI users and Codex-specific workflows.

Gemini CLI

[planned]

Useful for Google/Gemini users and long-context workflows.

Aider

[planned]

Popular open-source terminal pair-programming workflow.

OpenCode

[experimental]

Useful for open-source agent workflows and multi-session patterns.

Open model tooling

[experimental]

Selected provider-backed open models for low-cost tasks.

View supported agents

pricing

Credits that make token usage understandable.

Buy credits once. Use them across supported OpenAI, Claude, Gemini, and selected open models. See input, cached input, and output token costs before you delegate expensive work.

No subscription required during private beta.

See pricing

Credit balance

105.00

1 credit = $1

of managed model usage

Example model rates in credits per million tokens
ModelIn / 1MOut / 1M
Qwen3 Coder0.542.16
Claude Sonnet 53.7518.75
Claude Opus 4.86.2531.25

Live rates in credits per 1M tokens · full table

guardrails

Agent autonomy needs guardrails.

  • Per-task isolation
  • Scoped secrets
  • Explicit approval policies
  • Audit logs
  • Spending limits
  • PR-first review
Read security model

Request access to the private beta.

We are inviting developers and small teams who already use AI coding agents and want an always-on cloud environment for serious work.

Prefer more room? Use the full form.