Get started

Three commands and you're running.

LambAI runs on your own machine. You bring your key from us, and your model key. Nothing of yours is sent anywhere except to the model you choose.

Sign-in required — the download is for accounts, beta or otherwise.

1

Find your API key

It is on the screen you saw when you signed up — and it is the only time it is shown. Lost it? Rotate to get a fresh one; the old one stops working immediately.

Your keys →
2

Download the binary

Linux, x86-64, one file. No installer, no package manager, no dependencies to drag in.

3

Make it runnable and start it

Two commands. The agent opens a terminal prompt and works in whatever folder you launch it from.

On the command line

Replace lk_YOUR_KEY with the key from your signup — the whole thing, including the lk_ part.

  1. Download it

    curl -H "Authorization: Bearer lk_YOUR_KEY" \
      https://lambai.dev/api/download -o lambai

    Signed in already? Just use the button at the top of this page.

  2. Make it runnable

    chmod +x lambai
  3. Start it

    ./lambai

    It works in the folder you launch it from — that folder is the project. Run it somewhere you would be happy to have an assistant editing files.

    The first time, it asks for your licence key — every copy needs one. Paste the key from your signup, the whole lk_… value. It is remembered, so it will not ask again; it is re-checked on each start, which is what makes a revoked key stop working. Run it once while online: a key that cannot be checked is not saved.

  4. Or register the key yourself

    /license lk_YOUR_KEY

    The same thing, from inside the agent, and the way to move a key to another copy or replace one that was rotated. It sends your key and nothing else — no code, no prompts, no paths. See what we send and store.

    Nothing to type into — a script, a container, a service? Set LAMBAI_LICENSE_KEY or pass --license lk_… instead.

Give it a model

LambAI has no model of its own — that is deliberate, and it is why we can stay free while we are in beta. Point it at whatever you already use.

A hosted model

Any OpenAI-compatible endpoint. Put the key in the environment and LambAI will find it:

export DEEPSEEK_API_KEY=sk-...

Inside the agent, /key sets it for you, and /model picks which one to use.

Or entirely local

With Ollama running, your prompts and your code never leave your machine. You still need a free licence key to start — every copy does — but with a local model the key is the only thing that ever reaches us: one check on startup. The most private option, and the slowest.

ollama serve

Things worth knowing before you start

Linux for now
The binary we serve is Linux on x86-64. macOS and ARM builds are not ready. We would rather say so here than have you find out at the first command.
This is a beta
Free while we are testing it with real people, no guarantees, and things will change. If we ever introduce paid plans, beta users hear first and keep a discount.
The agent can edit your files
That is the point, and it is also the risk. It is confined to the folder you run it in, and it asks before writing until you tell it otherwise. Start it somewhere you can afford to be surprised.
Your key stays yours
We store only a hash of it, so we cannot show it back to you — and a copy of our database would not hand anyone a working key. Lose it and you rotate; that is a feature, not an inconvenience.