Skip to content

Run coding agents
in YOLO mode
on a sandbox that isn’t
anyone’s laptop

Alecoda gives every project its own isolated cloud sandbox with a live preview. Your Claude Code plugins install unchanged, and the agent runs on the model you pick for each task.

We’ll email you once, when it’s ready.

  • Claude Code plugins, unchanged
  • 9 model families
  • Hosted, or run on your infrastructure
A laptop running a coding agent with approvals off. The agent is reading the SSH key. Waitlist open. First version in 2027.

Your agents already have
the keys to every laptop

Within the agent’s reach on a laptop

~/.ssh/id_ed25519
SSH key
.env
Production secrets
~/.bash_profile
Personal API keys
~/
Everything else in your home folder

Permission prompts are miserable, so most developers turn them off. After that, the agent runs with the laptop’s full access: the .env files, the SSH keys, the home directory.

The API keys sit in each person’s .bash_profile. Nobody can see who is using which model or what it costs. When someone leaves, their key may leave with them.

“…our ‘governance’ for it was essentially nothing.”

A security and infrastructure owner on a team of 60+ engineers using Claude Code, writing on dev.to, June 2026

The session leaves the laptop
and lands in a sandbox

Same agent, same plugins, same conversation. Every command now runs inside one project’s sandbox, far from the files on your machine.

The agent session moves off the laptop, which keeps its keys and files. It lands in a sandbox for the ferrylabs checkout api project with 4 GiB of memory and 2 CPUs. Three plugins install, DeepSeek is chosen as the model, and the preview goes live on localhost:5173.

Three steps from
laptop to sandbox

Let the agent runevery command it needsJust never onanyone’s laptop

What the agent gets on day one

  • Shipped

    One sandbox per project

    Each project runs in its own isolated cloud Sandbox, capped at 4 GiB of memory and 2 CPUs by default. The agent can run any command there, with a live preview of your dev server beside the conversation.

    • 4 GiB memory
    • 2 CPUs
    • Isolated
  • Shipped

    Files that survive the sandbox

    If a Sandbox is lost, a new one starts on the same files automatically. Idle projects go Dormant and use no compute until you come back.

  • Shipped

    Your Claude Code plugins, as they are

    Install the Skills and Subagents your team already wrote, from any git source. Each install is pinned to an exact revision, and you're told when the source drifts from it.

  • Shipped

    Plugins that can't run code of their own

    Plugins load as data, with no hooks and no locally spawned MCP servers. A plugin can do nothing the agent couldn't already do inside its sandbox.

  • Shipped

    Long jobs in the background

    Hand a long task to a Subagent and carry on in the Session. Its work stays out of your main context.

  • Shipped

    The model you choose for each task

    When one vendor caps you or changes a policy, switch models and keep working.

    • Claude
    • GPT
    • Gemini
    • DeepSeek
    • Kimi
    • GLM
    • Qwen
    • Mistral
    • StepFun

Built next: control for
the person who pays the bill

These are on the roadmap and not shipped yet.

Join the waitlist
  • Spend you can see

    Usage and cost per developer and per project.

    Coming
  • Limits you set

    Caps and credits per person, so one heavy session can't drain the team's pool.

    Coming
  • Access you can revoke

    Roles and permissions per project, and nothing left behind on a laptop when someone leaves.

    Coming
  • Work you can hand off

    Pass a live Session to a teammate.

    Coming
  • GitHub integration

    Clone a repo into a project and work on it with the gh CLI.

    Coming

Questions a CTO asks first

Something else? Email hello@alecoda.dev.

We already use Claude Code locally. Why move?

Keep your plugins. They run here unchanged, in a sandbox that isn't your laptop, on any model. You don't have to choose between the workflow you built and keeping agents off your machines.

Isn't a devcontainer or a separate Linux user enough?

For one developer, often yes. A team ends up with each person maintaining their own sandbox, and no single place to see who is running what, on which model, at what cost.

Can we run it ourselves?

Yes. You can run Alecoda on your own infrastructure under our license, so your code never leaves it. We'll publish the license terms before launch.

Where does our code live on the hosted version?

On our servers. Each project runs in its own isolated Sandbox, and its files are kept separately from the Sandbox so they survive it. If your code can't leave your infrastructure, run Alecoda yourself instead.

What will it cost?

There's no pricing yet.

When can I try it?

In 2027. Join the waitlist and we'll email you when it's ready.

Get your team’s agents
off its laptops

Join the waitlist and we’ll email you when the first version is ready to try.

We’ll email you once, when it’s ready.