Nouride

Self-hosted multi-agent runtime

Serious agents. Tiny runtime.

Your own AI assistants, on your own machine, answering in Telegram, Discord and WhatsApp. One binary, one directory, one browser tab to set it up.

Start here What is this?

Install on Linux or a Raspberry Pi

curl -fsSL https://get.nouride.com/install.sh | sudo sh

Read it first — it runs as root on a machine nobody has logged into yet: curl -fsSL https://get.nouride.com/install.sh | less

What it actually is

A daemon you run. Not a service you sign into, and not a wrapper around somebody else’s assistant.

Runs on your hardware

One compiled binary and one directory. No Docker required, no Python, no GPU, and no Postgres, Redis or vector database — the whole install is a SQLite file next to the executable.

Several agents, each its own

Every agent has its own identity, model, tools and permissions, kept as Markdown and TOML you can read, edit and put in git. They can hand work to each other.

Answers where you already are

Telegram, Discord and WhatsApp, plus a dashboard and a terminal. The same agent, the same memory, whichever one you reach for.

Asks before it acts

Anything that touches the machine goes through an approval you answer from the chat you are already in. Privilege escalation, recursive deletes and credential paths can never be auto-approved.

The dashboard

Set it up, watch a turn as it happens, answer approvals, read the log. Served by the daemon itself — there is nothing else to deploy.

The dashboard overview
A conversation with live turn steps
The approval queue
Scheduled work

Five minutes, start to finish

  1. 1

    Run the installer

    It creates a service account, installs the daemon beside its dashboard, writes a systemd unit and starts it.

  2. 2

    Take the first password

    Printed once, in the log, on the first boot against an empty database. You are made to change it at first sign-in.

  3. 3

    Finish in the browser

    Point it at a model provider, give the first agent a name and a personality. Nothing needs a terminal from here.

  4. 4

    Connect a chat app

    Paste a bot token, or scan a QR for WhatsApp. Then say hello from your phone.

Read the install guide

Measured, not claimed

Every figure here was taken on real hardware, and the method is written down beside it.

64 MB
RAM it boots in 48 MB dies, 64 MB reaches ready
~95 MB
installed on disk binary and dashboard together
0.7%
of one core, idle it waits on the network almost always
0
services to run beside it no Postgres, Redis, vector DB or GPU
System requirements, with the measurements →

Where it answers

Several bots on one platform is a supported shape — one agent per token, each with its own name and personality.