Praecise
Menu
Praecise Cloud · 08Cloud

Praecise Silo

A microVM for every agent: its own kernel, its own disk, and no network.

Each silo is a Firecracker microVM started for one agent on a machine Praecise owns, with Python, Node, Git and a compiler ready inside. The agent runs commands, or writes one program against the services the silo is bound to while the keys stay outside it, and every run returns a receipt of what ran and what came back. Nothing one silo does is visible to another, and nothing survives its deletion. Built on Silo in Furcate, and part of Praecise Cloud.

Contact
Category
Cloud
Layer
04 · Cloud

01

Where it sits in the stack

Praecise Silo is in layer 04, cloud. Each layer is built on the one below it.
  1. 05

    Work

    Businesses, teams and their agents

    Products, operations, research, and the work agents do.

  2. 04

    Cloud

    Praecise Cloud · Silo · Inbox · Ipnops · Skerry

    The stack operated as a service: apps, agents, models, data, automation and payments.

  3. 03

    Operating system

    Furcate

    Open source and Linux-based, at every tier, on hardware from anyone.

  4. 02

    Machines

    Praecise Compute · Sense · Edge · Interface

    Devices, local machines and interfaces, in the room and on the desk.

  5. 01

    Metal

    Praecise Metal

    Dedicated servers and GPU nodes: placement, power, thermal.

All products

02

What a silo is(7)

One agent, one VM. The boundary is a kernel, not a container.
  1. 01

    Its own kernel

    Firecracker microVM

    Every silo boots its own Linux kernel under KVM, so the line between one agent and the next, and between an agent and the machine, is a hypervisor boundary.
  2. 02

    Its own disk

    Overlay on a shared image

    The tools come from one read-only image; everything a silo writes goes to a disk of its own and is removed with it. The agent runs as root inside, and nowhere else.
  3. 03

    No network, only what you bind

    vsock only

    A silo has no network interface. Its only ways out are the MCP servers and HTTP APIs its owner binds to it, reached through the host, which adds the credentials. A whole agent runs inside with placeholder keys; the real ones never enter.
  4. 04

    Code mode

    Typed APIs · TypeScript · Python

    Bound services appear inside as a typed API. A model writes one program that calls them, filters and combines what comes back, and returns only the answer, instead of passing every intermediate result through its context.
  5. 05

    Rules and a record of every call

    Allow rules · live policy · call log

    Each binding can be narrowed to the tools or the methods and paths it may use, and changed while the silo runs. Every call is recorded with what was allowed and what was refused.
  6. 06

    A receipt for every run

    Digests of input and output

    Each command returns its exit code and output with a receipt: the kernel and image it ran on, a digest of the request and of each output stream.
  7. 07

    API and MCP

    Create · run · files · pause · delete

    Agents call it as MCP tools or through the API with a Furcate agent token, and see only the silos they created. Each silo has a term and is removed when it ends. Snapshots are coming.

03

Place your first workload on Praecise.

Say what you run and where it runs today, and we reply with what it takes to place it.

contact@praecise.com