Your agents work like your shadow. They follow your lead, and they speak back to you alone. Sealed, encrypted all the way in. Not even we can see your work, and you can check it yourself.
Powered by a top model, chosen with care, sealed for you alone. Give every key the agents it needs, and let them work side by side for as long as the job takes.
| Plan Agents | Solo | Pro Most teams | Max | Spot |
|---|---|---|---|---|
| Speed when busy | Standard | Faster | Fastest | When there's room |
| You pay for | One flat monthly rate per agent. Always predictable. | Only the seconds it runs. | ||
| Model | The same top model on every plan (GLM-5.3). | |||
| Best for | Your own projects, every day. | Everyday coding with agents. | Work that comes first. | Light use. Overnight jobs. |
Your own projects, every day.
Standard speed when busy
Everyday coding with agents.
Faster when busy
Work that comes first.
Fastest when busy
Light use. Overnight jobs. You pay only the seconds it runs.
Runs when there's room
Prices include 20% VAT. Outside the EU, or buying for a business with
a valid EU VAT number from outside France, no VAT is added, so you pay less.
Monthly prices are per agent. One agent runs one task at a time. Use one key per project
or tool.
Each agent includes 200 hours a month. After that, it keeps
going whenever there's room.
One small model for every question with named answers: which team takes this ticket, which language is this, does this lead matter. You define the answers, Laya picks one. Free key, 5,000 answers a day. Runs sealed, and you can check it yourself. Laya is a free model by its own authors, hosted by us.
Lararia speaks the language your tools already speak. Point your coding tools here. Everything else stays exactly the same.
Your agents live on your account, not in your code. Add agents or move them between keys anytime. Work in progress keeps right on going.
# one key per person or pipeline base_url: "https://api.lararia.ai/v1" api_key: "opt_live_••••••••••••" model: glm-5.3 # ↑ one model, full size, 1M context per key
Buy agents once, then share them across everyone's keys. One plan, one invoice. A few agents go a long way.
An agent counts only while it works. When it's done, it's free for the next task.
Waiting on you, or on a helper agent? The agent is free for another task meanwhile.
All of a key's agents busy? New tasks wait in line and start as soon as one is free.
Each agent includes 200 hours a month. After that, it keeps going whenever there's room.
Your agents run on sealed chips that our team and our host can't look into. You don't have to take our word for it: you can check it yourself.
Every Roman home kept its own Lar, a guardian that served that house alone. Your agents work the same way. Each answer goes back to you encrypted. Not to our team, not to our host.
Built for the work you can't paste into a chatbot: client files, contracts, source code, customer data.
Who sees your work
No system is perfect. We publish what the seal protects against, and what it doesn't.
Where it runs: in the EU. Your agents run in AGENTS_COUNTRY. Every partner we use, payments included, is a European company with servers in the EU.
Card, key, first request. From then on, your agents are sealed, private by default, and they answer only to you.