Questions about LStack

LStack is an AI development team for your projects. The app runs on your computer and puts the AI coding tools you’ve set up to work. You describe the work in tickets (each one a feature, fix, or improvement), including how you’ll know it’s right. AI can help you write them. LStack coordinates the coding, your chosen review, and any corrections, then picks up the next ready task. When a ticket needs you, it says so.

It isn’t a new AI model, and it doesn’t promise finished, production-ready software. You still decide what’s good enough.

An AI coding tool such as Claude Code writes the code; LStack works a layer above it. You direct the project through tickets and their acceptance criteria. LStack hands each ready ticket to the AI coding tool, sends the changes through the review you chose, carries feedback into corrections, and moves on to the next ready task, keeping a record on each ticket. You manage priorities and decisions on the board instead of running each coding session yourself. It doesn’t replace your editor or Git.

Today the AI coding tool LStack runs for you is Claude Code. Support for more tools is Planned.

Founders, product managers, and developers who want to direct software work through tickets instead of supervising one AI chat after another, and anyone experimenting with local or self-hosted models. Most of the work happens in the dashboard, not the command line, though you’ll still want someone who can set up the AI tools and judge whether the code is right.

You write tickets and move the ones you’re ready for to To Do. When a suitable AI coding tool is available and allowed to work on that project, LStack gives it the next ready task. Finished work moves to Needs Review, and accepted work lands in Done.

With automatic review turned on, a separate AI review checks the changes before they’re accepted. If something goes wrong, LStack can retry some problems or pause the ticket with a reason you can read on it. It doesn’t quietly mark work as done.

Only if you choose that. Each project decides how finished work is checked: you review it yourself (manual review), a separate AI review checks it (automatic review), or the project opts in to finishing without a separate review step (direct completion).

An automatic review can use the same model and setup as the coding or a different one, for example a local model writing and a cloud model reviewing. On the ticket you’ll see the review’s decision, a short summary, and any reasons for rejecting the work. An AI review doesn’t guarantee correct or secure code, so check the result before you rely on it.

Yes, as long as what it needs stays available: your computer on, LStack and Docker running, and your AI services reachable with capacity to spare. Ready tasks keep getting picked up while you’re gone. If the computer sleeps or a service is unreachable, the work can’t continue until it’s back.

A task that needs a decision waits on its ticket, with its history. Open it to see what’s needed, and once you’ve dealt with it, move the ticket back to To Do.

It describes where LStack itself runs. The dashboard, your projects and their files, the task history, and any project you set up to run all live on the computer you install it on. It doesn’t mean every model has to be local, or that nothing leaves your machine: a cloud AI service receives the instructions and code it’s sent for each task.

Yes. You pick the AI service for each AI coding tool you set up: all local, all cloud, or a mix, such as a local model writing code and a cloud model reviewing it. Only supported services work: today that means ones offering the Anthropic Messages API, including local model servers and hosted endpoints.

Providers and models lists what’s supported and how to set it up.

Only if you turn it on. By default LStack is reachable only from the computer it runs on. Optional LAN access lets other devices on a network you trust sign in. Basic LAN mode isn’t encrypted, so keep it to trusted networks. LAN access covers the details.

Anyone on a supported platform can install and use LStack today, with no invitation or waitlist. It works on real projects, but it’s still evolving: features and workflows can change between releases, and it isn’t meant for production-critical work yet. Review the code AI produces before you rely on it.

Run one command in a terminal to install the latest release. Installing doesn’t start LStack: run lstack start, open the dashboard, and follow the setup steps to connect your AI tools and create your first project.

• Apple Silicon macOS, or x86-64 or ARM64 Linux. Windows is not supported; WSL2 is a workaround.
• Docker, for the AI coding tools and for projects that run in containers. The dashboard runs without it; stack mode also needs Docker Compose v2.
• A supported AI service, local or cloud
• No account needed for local use

The installation guide has the full steps.

docs.lstack.ai covers setup and troubleshooting. Report bugs in the issue tracker. For anything else, use the contact page. Security concerns go there too, so they stay private; please don’t post them publicly.