Anh Tuan Labs

Trạm by Anh Tuan Labs

Run coding agents
in isolated sandboxes.

Anh Tuan Labs is an independent software studio. Our first product, Trạm, is a self-hosted workspace that gives every AI coding task its own Docker sandbox, so your source repository stays untouched.

Meet Trạm

Trạm is in development. Not yet publicly available.

01 / Current product

Trạm

In development · v0.1
Self-hosted. Not yet publicly available.

A self-hosted control plane for AI coding agents. Every task gets its own thread and its own Docker sandbox, so several agents can work at once without touching your source repository.

Trạm thread view: a conversation where an agent adds rate limiting to a checkout API, with the thread list on the left and the code diff on the right.
Thread view with sample data.
01

Isolated by default

Each thread clones your repository into its own Docker volume. The agent works on that copy and never writes to your original checkout. CPU, memory, and process limits are configurable.

02

Bring your own agent

Run existing command-line coding agents, or connect your own through a small newline-delimited JSON protocol.

03

Review what changed

Open the diff, browse changed files, and run commands in the sandbox terminal before you decide what to do next.

04

Dev services in the sandbox

Declare services in .tram/services.json. Trạm starts them inside the sandbox, checks their health, and links to them on your machine.

05

A coordinator for parallel work

Sen, the built-in coordinator, starts child agent threads, follows their progress, and asks for your approval before archiving them.

06

One CLI, optional web

A single tram binary runs a local daemon, an interactive CLI, and an optional web control surface. Several machines can join over a private VPN.

How a thread works

  1. Pick a repositoryChoose a local Git repository, an agent, and a prompt.
  2. Work in a sandboxThe agent runs in its own container with its own copy of the code.
  3. Review the resultRead the conversation, the diff, and any running services.
  4. Follow up or archiveSend another prompt, cancel, or archive the thread and its sandbox.

Runs on your machine and listens on localhost by default. Model provider sign-in tokens stay local and are never accepted as a Trạm login.

02 / The studio

Small by design.
Useful by intention.

Anh Tuan Labs is a small, founder-led studio focused on the everyday work of building software.

Trạm is our first product. It comes from a practical problem: running AI coding agents on real projects while keeping each change contained, visible, and easy to review.

We focus on AI-assisted development, developer workflow automation, and tools for building and operating software. AI is part of the toolkit. A better developer experience is the goal.

Clear behavior. Thoughtful defaults. Less friction.

03 / The person behind the studio

Built by a developer,
for developers.

Anh Tuan

Founder & Developer

I'm a full-stack developer working with React, Elixir/Phoenix, and AI integrations. Trạm started as a tool for my own projects. Anh Tuan Labs is where I turn that kind of tool into something other developers can use.

ReactElixir / PhoenixAI integrationsDocker

04 / Get in touch

Interested in Trạm?

Email to follow its development, or tell us
how you run coding agents today.