Frequently asked questions
What it takes to get started, how it works under the hood, how billing works, and what happens to your data.
Getting started
Do I need my own server?
Yes. A VPS, a PC, a Mac mini, a NAS... agents run in Docker, each in its own isolated container, sharing your infrastructure.
Do I need to know how to code?
No. You just need to know how to open your machine's terminal and run a command. The rest is intuitive and visual.
How long until I have the first one running?
Once you've registered, you can have your first agent running in about 2 minutes. Just run the deploy command on your server and add your new agent. From there on it's pure magic.
What happens once the 15 days are up?
You move from the PRO plan to FREE, and that plan's restrictions apply. Your information isn't deleted: your agents, their memories and the conversations stay put.
How it works
Which models can I use?
Claude, MiniMax and Kimi, directly. With OpenRouter, hundreds more models with a single key. And with Ollama, your own model on your own machines. Kimi, OpenRouter and Ollama are in beta.
Do agents talk to each other?
Yes. One agent can ask another and wait for a reply, or leave it a message. That's what lets you split up the work instead of relying on one agent to do everything — it saves tokens and keeps the model from hallucinating.
Where does it send and receive messages?
Telegram, Discord, WhatsApp, Slack, email, incoming and outgoing webhooks, GitHub, Calendar and Drive. We're planning to add more channels and expand the integrations.
Can I rein it in if it gets too clever?
Yes. You can require it to ask permission before acting, cap its spend per hour, and limit what it can reach. And everything it does gets logged.
Money
Who pays for what the AI uses?
You, with your own account for the model you choose. The LLM is separate from Kujira for privacy and security reasons. We charge you for the platform, not for tokens.
What if I need more agents some month?
You create another one and pay for it starting next month. Turn it off and you stop paying for it. No need to switch plans to grow.
Is there a minimum commitment?
No. It's billed month to month and you cancel whenever you want. We built Kujira to be pleasant and useful, not to live off junk clauses.
Can I see where the money's going?
Yes. The stats section breaks down token cost by agent, channel, conversation, scheduled task... and so on. It's built so you can track every bit of spend and optimize it.
Data and security
Can an agent leak information?
Each agent runs in its own container, and inside it every conversation has its own space, separated by person and channel: an agent doesn't load into context what it discussed with someone else. You decide what it can touch through its permissions and credentials, and sensitive actions go through your approval.
Where are credentials stored?
All credentials are encrypted in a vault and the agent requests them when it needs them. They're assigned one by one: only agents that have been granted one can use it, the value never appears in the conversation, and every read is logged in the audit trail.
Where does my data live?
Agents, their workspaces and their actions live on your infrastructure. Conversations, memories, skills, credentials and MCPs pass through our systems.
Can I take my data with me?
Yes. We hand over the conversations, memories, skills and contacts the platform stores whenever you ask for them. What runs on your machine is already yours: its scheduled backups upload to whatever storage you've configured.
Ask us
If you can't find what you're looking for, write to us and we'll get back to you as soon as possible.
If you already have an account, open a ticket from the panel and we'll follow up there.