Imperal Cloud — the AI Cloud OS where the AI runs inside the kernel, not on top

Imperal Cloud is the AI Cloud OS where the AI runs inside the kernel — not on top of it. Other platforms ask the AI to behave: tool-use protocols, agent libraries, plugin stores. We built the kernel that contains it. Every action the AI takes passes through a federal contract enforcing authorization, isolation, audit, and anti-hallucination — before your extension code runs.

The world's first complete implementation of ICNLI — Infrastructure Contextual Natural Language Interface, the open protocol we authored and published under CC BY-SA 4.0 that maps natural-language intent to context-aware infrastructure actions. Anyone can implement ICNLI; we are the first who shipped it as a deployed multi-tenant cloud OS.

Webbee 🐝 is its native agent — and it's two more firsts in one.

First cloud-modular AI agent. Every other AI agent today is a closed monolith with a bolted-on plugin store owned by one vendor. Webbee is the first AI agent that gains capabilities the same way an OS does — through federally-contracted extension packages authored by anyone, installable per-user, runtime-dynamic, LLM-provider-agnostic.

First proactive AI agent by design. ChatGPT, Claude, Gemini — all reactive. They wait for the user. Webbee is the first AI agent where proactive behavior is structural: ambient skeletons keep it aware, schedules let it act on time, webhooks let it react to events, and the platform owns chat-injection so the agent can speak first when something changes. No polling loops; no author plumbing.

You write one function. The OS handles authentication, per-user isolation, RBAC scopes, action-type confirmation gates, audit ledger, retries, timeouts, cancellations, long-running task delivery, multi-tenant safety, anti-hallucination invariants, and chain orchestration across extensions. The same way a Linux dev doesn't write their own network driver — you don't write your own AI-safety layer.

Welcome to Web 3.0.3 — three patches past the hype.

→ For skeptics: Why this is the AI Cloud OS — the six-property test, named-alternative scoring, engineering receipts, falsifiable claim of priority by ICNLI.

→ For builders: What we built that nobody else has — 20 platform-and-agent surfaces with zero precedent.

→ Why Webbee is unique: Cloud-modular + proactive by design — two distinct firsts neither competitor can claim.

💰 Build extensions, earn real money

Imperal Marketplace is a real revenue surface. Every user action through your extension generates token flow — your developer_share accrues in real time, withdrawable to USD via Stripe Connect.

Pick your starting line

What you can build

Install in one line

pip install imperal-sdk

from imperal_sdk import Extension, ChatExtension, ActionResult
from pydantic import BaseModel, Field

class GreetParams(BaseModel):
    name: str = Field(description="Person to greet")

ext = Extension(
    "hello-world",
    version="1.0.0",
    display_name="Hello World",
    description="Demo extension that greets people by name with a friendly message.",
    icon="icon.svg",
    actions_explicit=True,
)

chat = ChatExtension(ext, tool_name="hello_world", description="Hello World — friendly greetings.")

@chat.function("greet", action_type="read", description="Greet someone by name with a friendly message.")
async def greet(ctx, params: GreetParams) -> ActionResult:
    return ActionResult.success(
        data={"text": f"Hello, {params.name}! 🐝"},
        summary=f"Greeted {params.name}",
    )

That's a real, working extension. The LLM will call greet(name="Alex") when a user types "say hi to alex".

Built on hard guarantees

Install / upgrade

pip install --upgrade imperal-sdk

For per-release migration notes see the SDK changelog.

All pages on this site

A complete map of everything documented here. Pick anything to jump in.

🚀 Getting Started

🧠 Core Concepts

🛠️ Guides

📐 SDK Reference

📚 Recipes

📖 Reference

🎨 Components showcase