Your AI. Your key. Your record.
Tell it things as they happen. Months later, ask. It answers from what you actually wrote, keeps working while you sleep, and shows you every step it took.
Here is somebody's record of a decision they are in the middle of. Ask a real agent about it, right there. Nothing to sign up for.
Free and open source. Nothing to install, nothing to fill in. Mac, Linux and Windows too.
The problem
You explain your situation. It helps. Next week you explain it again. The context you built up is gone, the decision you made is gone, and the thing you were promised is gone with it. So the work stays shallow, and nothing ever compounds.
Lamdis keeps it. Everything you write, everything it answers, every step it takes, in one record that is yours and stays put. That is the whole difference, and it changes what an assistant is for.
What that gets you
Drop in a quote, a diagnosis, a school date, a policy number, whatever crosses your desk. In March you write it, in September you ask. It tells you, and tells you which day you wrote it. No folder to keep, no note to find, nothing to re-explain.
Every quote given, every price agreed, every date committed to a customer. Ask before you answer them. Tell it to watch a thread and it flags the contradiction itself: you told the bank mid-November, and this vendor now starts six weeks out.
Every action an agent takes is written down and signed: what it read, what it changed, which tools it called, what it cost. Access is granted by a person and revocable. You can hand somebody a thread at the depth you choose, and nothing else.
Bring your own key
Everyone else picks the model for you and charges a subscription on top. Lamdis takes your key and talks to whatever you point it at: hundreds of models through OpenRouter, or one running on your own hardware for nothing. Every run in your record shows what it cost, in cents, beside what it did.
The key sits on your machine and goes to your provider. We are not in the middle of it, we do not mark it up, and there is no subscription to cancel.
Point it at Ollama or vLLM on your own hardware and the whole thing runs locally: your record, your agent, your model. No key, no bill, nothing leaving the machine.
| Model | Per million tokens, in / out | A typical question |
|---|---|---|
| OpenAI GPT-5.6 Luna default | $0.20 / $1.20 | less than a penny |
| Anthropic Claude Sonnet 5 | $2.00 / $10.00 | a few cents |
| Anthropic Claude Opus 5 | $5.00 / $25.00 | a few cents more |
| Your own hardware | free | free |
Switch between them mid-conversation. No key yet? Write to us and we will send you one with credit on it.
While you are not there
In plain words: what should it do when something new arrives, and what should it stop and think about each morning. From then on it works on its own, and everything it does is written into the thread for you to read when you get back.
Something lands in a thread and it checks it against everything else you have written, then says what does not add up.
Give it an hour. A morning pass that asks what needs you today; an evening one that looks for what is worth sleeping on.
When the call is yours it asks, with options, and waits. It will not commit you to anything and it cannot give anybody access.
Leave one running on your laptop with lamdis agent and direct it from your phone. It works where you let it, asks before going further, and writes down everything.
Getting started
Open it and you are in a thread, typing. No account, no card. Add an email later if you want it on your phone too.
Mac and Linux. Windows: irm https://lamdis.ai/install.ps1 | iex
They read and write the same record, so two agents stop repeating each other's mistakes.
Reads the repository, makes the change, runs your tests, and files what it did.
Straight answers
ChatGPT forgets between sessions and picks the model for you. This keeps everything you have ever told it, runs on whichever model you choose at cost, and can act while you are away.
No. Point Claude Code or Codex at Lamdis and they work in the same record. It is the memory underneath, not a replacement for the thing you like.
In one file. On your computer, or on ours if you would rather not run anything, and you can take that file and run it yourself whenever you like.
Your provider, and nobody else. It stays on the node you run, we never proxy your requests, and a local model needs no key at all.
It reaches only the sites and tools you allowed, refuses credential stores outright, stops to ask when a choice is yours, and writes down every single thing it did.
The software is free and open source. You pay your model provider directly at their price. A light day is a few cents; a local model is nothing.
Ready?
app.lamdis.ai — or keep the whole thing on your own machine:
Mac, Linux and Windows. Free and open source, on GitHub.