It comes to you
Start with a brief of your inbox, upcoming plans and the decisions that need you. Less babysitting, not more.
Lares is an open-source platform for running a small team of AI agents inside your business. They read the inbox, keep the calendar, draft the follow-ups, and ask before anything goes out. Your server or ours, your keys, your rules.
No dashboard to log into. The morning brief arrives in the chat you already use. You answer in one tap. The agent does the rest and keeps a receipt of what it did and who said yes.
Morning. Two things need you. The rest is handled.
Follow-up to Kari after yesterday's meeting. Draft ready.
Thursday has two overlapping calls. Move the 14:00?
Send it. Move the 14:00.
Sent to Kari. 14:00 moved to 15:30, both invited. Receipt saved.
Start with a brief of your inbox, upcoming plans and the decisions that need you. Less babysitting, not more.
Review an action before it goes out. Set what each agent may do, and when it must ask.
Your server, your keys, your data. One official, private place for the work people are already pasting into chatbots.
Choose a purpose, give it a name, and set its access. Keep its work and permissions in one place.
Install on your own server today. Your keys, your rules, your agents. Open source, Docker, one command.
Install guideA standard installation on a server we run and update, in a region you choose. Opens when enough of you ask.
Set up and cared for by us, on our servers or inside your own infrastructure. Starts with a conversation.
You want the manual work gone, not a new system to learn. Tell us what it is. We'll say honestly whether lares can do it yet, and what it would take.
Email opens in your mail app. Booking opens a small calendar window.
What agents learn is Markdown in git, readable in any editor. Dated facts sit in a database and every one can be exported.
An agent may add to what it knows. Changing or forgetting something needs your yes. Nothing read in an email can become a memory.
Dated monthly releases. A backup before every update, a fix shipped with every breaking change, one command to roll back.
Nightly, encrypted. A monthly restore drill. Export everything, or erase one person's data on request.
Bring your own model keys. A gateway on your server maps each job to a model, and every agent has a daily spend cap.
Runs on your own server, or ours in a region you choose. Outbound traffic goes through an allow-list. No telemetry.
Several people on one installation, each with private and shared areas. Mail and calendar on Google or Microsoft 365. Briefs, drafts and approvals in Slack, Telegram or web chat. Tracks for what matters, without a CRM.
describes the coming release
Whatever you give it keys for. The engine does not bundle a model or a subscription.
It stays on the server you chose. The model provider sees what you send it, and only that. No telemetry.
Self-host is free and open source; you pay your model provider. Cloud and managed pricing is still being set.
The Roman lares were the spirits that kept a house running and kept bad things out. They were tended daily and never worshipped; the shrine was a practical fixture in the wall, a config file. A Greek daimon became a Unix daemon. Lares are trusted daemons in your house.