Skip to main content
How we run your AI assistant in the cloud.

​
Overview

EasyClaw uses a two-tier architecture:
  1. Public API — A REST API handling authentication, session management, and credential storage.
  2. Fleet Agent — A container orchestrator running on dedicated servers, managing private containers where your AI assistant actually runs.
┌─────────────┐     ┌──────────────┐     ┌────────────────────────┐
│  Your phone  │────▶│  Public API   │────▶│  Fleet Agent            │
│  or browser  │     │              │     │                        │
└─────────────┘     └──────────────┘     │  ┌──────────────────┐  │
                                          │  │ Your Container   │  │
                                          │  │ ┌──────────────┐ │  │
                                          │  │ │  ZeroClaw     │ │  │
                                          │  │ │  (your AI)    │ │  │
                                          │  │ └──────────────┘ │  │
                                          │  └──────────────────┘  │
                                          └────────────────────────┘

​
The public API

The API handles everything that isn’t running your AI:
  • Authentication — User sign-up, login, session management
  • Credential storage — API keys encrypted before being stored in the database
  • Session lifecycle — Spawning, freezing, restoring, and killing containers
  • Usage tracking — Token counting, cost calculation, and plan enforcement
  • WebSocket proxy — Tunnels real-time connections from the browser to your container
The API never processes your AI conversations directly. It delegates all AI work to your private container.

​
The Fleet Agent

The Fleet Agent runs on dedicated servers and manages the lifecycle of containers:
ActionWhat it does
SpawnCreates a new container from a template image
FreezeSuspends a container to free resources
RestoreResumes a frozen container
KillDestroys a container permanently
Communication between the API and Fleet Agent is authenticated with a shared secret.

Fleet Agent deep dive

Container lifecycle, DNS, and reverse proxy management.

​
Private containers

Each user gets their own container — a lightweight, isolated Linux environment. Containers are cloned from a template image using copy-on-write, which makes them spin up in seconds. Inside each container:
  • ZeroClaw — The AI agent framework
  • Messaging bridges — WhatsApp, Signal, Telegram, iMessage, Email
  • Browser — Headless browser for web browsing and automation
  • Tools — File system, shell, text editor, and more
  • Persistent storage — Files, memory, and conversation history survive restarts

Containers deep dive

Template images, instant cloning, persistent volumes, and freeze/restore.

​
ZeroClaw

ZeroClaw is the core AI agent framework that runs inside every container. It handles conversation context, tool execution, memory, channel bridging, and model provider connections.

ZeroClaw deep dive

Configuration, autonomy levels, tool system, and memory.

​
Data flow

Here’s what happens when you send a message on WhatsApp:
1

Message arrives

Your WhatsApp message is received by the messaging bridge running inside your container.
2

ZeroClaw processes it

The bridge forwards the message to ZeroClaw, which loads your conversation history and decides how to respond.
3

AI generates a response

ZeroClaw sends your message (with context) to your chosen AI provider. The AI may decide to use tools — browse the web, check your calendar, etc.
4

Response is sent back

The AI’s response is sent back through the messaging bridge to WhatsApp.
The entire flow happens within your private container. The public API is not in the message path — it only manages container lifecycle and authentication.

​
Infrastructure stack

ComponentHosted on
FrontendCloud edge network
Public APIManaged cloud platform
Fleet AgentDedicated servers (EU)
ContainersDedicated servers (EU)
AI FrameworkInside containers
DatabaseManaged PostgreSQL
DNSEdge DNS provider
Reverse ProxyDedicated servers
Object StorageCloud storage