SOUL.md v1.0.0 · Production-Ready · Node 20 + TypeScript

Your First
AI Employee

AIKoClaw spins up self-destroying agent sandboxes, remembers your full context, and executes real tasks — across every messaging app you already use.

aikoclaw · deploy.sh

$ curl -fsSL https://aikoclaw.io/install | bash

✓ AIKoClaw v1.0.0 detected

✓ Docker runtime ready

✓ Kubernetes cluster connected

✓ Telegram bot token verified

⟳ Spawning root agent @ @aikoclaw_root_bot▊

Deploy AIKoClaw Free Live Dashboard

Privacy by design · Self-destroying sandboxes · Zero trace sessions

3 agents active

⚡ Task complete · 1.2s

🔒 Sandbox destroyed

Capabilities · SOUL.md Powered

Superpowers, Built In

Every AIKoClaw agent runs with the full SOUL.md spec — lane queues, semantic snapshots, self-destroying sandboxes, and circuit breakers out of the box.

🧠

Persistent Memory · 4 Tiers

Redis ephemeral sessions, PostgreSQL episodic logs, Qdrant semantic vectors, and snapshot deltas. Your agent never forgets.

RedisPostgreSQLQdrant

🔥

Self-Destroying Sandboxes

Sub-agents spin up isolated K8s pods, execute tasks, then destroy themselves — zero session traces, zero residual data.

DockerKubernetesEphemeral

🔒

Privacy by Design

Vault-backed secrets, HMAC-signed A2A messages, circuit breakers, and budget hard-stops. Your cloud. Your rules.

VaultHMACZero-Trust

⚡

Lane Queue System

P0 critical to P3 low-priority lanes with exponential backoff, dead-letter queues, and human-review escalation built in.

P0→P3RetryDLQ

📡

A2A Telegram Protocol

Agents communicate peer-to-peer over signed Telegram messages. Task delegation, snapshot refs, and TTL-bounded payloads.

A2ASignedTTL

📊

Real-Time Telemetry

Loki logging, Prometheus metrics, PagerDuty alerts. Lane depth, error rate, circuit breakers — visible in your dashboard live.

LokiPrometheusGrafana

SOUL.md Architecture

How the Agent Stack Works

Every deployed AIKoClaw instance runs this exact production-grade stack from your SOUL.md specification.

User Layer

💬 WhatsApp

✈️ Telegram

⚡ Slack

🟣 Teams

↓ A2A Protocol (HMAC-signed)

Root Agent

🦞 @aikoclaw_root_bot
Lane Router · Circuit Breaker · Snapshot Engine

↓ Spawn Protocol (K8s Pods)

Division Agents

📊 Finance Bot

⚙️ Operations Bot

📣 Marketing Bot

+ Spawn New

↓ Sub-Agent Sandboxes (Ephemeral K8s)

Sandboxes

🔲 Spin Up Pod → Execute → 💥 Destroy + Wipe

↓ Persist to Memory Tiers

Memory Tiers

⚡ Redis
Ephemeral / 24h TTL

🐘 PostgreSQL
Episodic / 90 days

🧠 Qdrant
Semantic / Vectors

📸 Snapshots
Delta / Audit

Agent Lifecycle · SOUL.md §7

From Message to Completed Task

Six precise lifecycle steps — spawn, execute, snapshot, persist, unregister, destroy.

Deploy @AIKoClaw Bot

Run the one-line install. Your root agent registers on Telegram, verifies all memory tiers, and announces via A2A broadcast.

Agent Learns Your Context

Share your division, budget limits, goals. Qdrant stores semantic vectors. Every interaction sharpens the context model.

Task Enters Lane Queue

Your message is classified P0–P3. Concurrency, retry policy, and timeout auto-apply. Circuit breakers protect downstream.

Sandbox Spins Up

A fresh K8s pod instantiates, inherits the agent soul, executes with real-time semantic snapshots every 5 seconds.

Result Delivered

Task output sent back via Telegram / Slack / WhatsApp. Snapshot delta stored to PostgreSQL and Qdrant for future learning.

Sandbox Destroyed

Container wiped. Session traces purged. A2A registry updated. Zero residual data. Ready for the next task instantly.

Live agent interaction

/spawn finance — budget cap $5,000, division: procurement 🚀

@aikoclaw_root_bot ✅ Pod finance-001 spawned · P1 lane · Budget circuit: $5,000 hard-stop set · Registered at agent.aikoclaw.local · Telegram linked. Ready.

Research server hardware options under $3k and book the cheapest option

@aikoclaw_finance_bot 🔲 Sandbox spinning up… Researching 14 suppliers… ✅ Found: Dell PowerEdge R350 at $2,847. Order placed. Snapshot saved. Sandbox destroyed. Zero traces. 🗑️

One-Click Deploy · SOUL.md §9

Get AIKoClaw Running

Choose your deployment method. From local Docker to full Kubernetes production — all from your SOUL.md spec.

bash · Quick Install

# One-line install — spins up full AIKoClaw stack
curl -fsSL https://aikoclaw.io/install | bash

# Then configure your bot token
aikoclaw init \
  --telegram-token YOUR_BOT_TOKEN \
  --domain your-domain.com \
  --enable-telemetry

# Start your first agent
aikoclaw spawn --name root --division headquarters

bash · Docker Compose

# Pull and run with Docker Compose
curl -fsSL https://aikoclaw.io/docker-compose.yml -o docker-compose.yml

# Set environment variables
export AIKOCLAW_VERSION=1.0.0
export TELEGRAM_BOT_TOKEN=your_token_here
export ENCRYPTION_KEY=$(openssl rand -hex 32)
export JWT_SECRET=$(openssl rand -hex 32)

# Spin up Redis + PostgreSQL + Qdrant + AIKoClaw
docker compose up -d

# Check status
docker compose ps
docker compose logs aikoclaw --follow

bash · Kubernetes

# Apply namespace and secrets
kubectl create namespace aikoclaw
kubectl create secret generic aikoclaw-secrets \
  --from-literal=telegram-token=YOUR_BOT_TOKEN \
  --from-literal=encryption-key=$(openssl rand -hex 32) \
  --namespace aikoclaw

# Deploy the full stack
kubectl apply -f https://aikoclaw.io/k8s/manifests.yaml

# Verify pods are running
kubectl get pods -n aikoclaw

# View live logs
kubectl logs -f deployment/aikoclaw-root -n aikoclaw

bash · Helm + ArgoCD

# Add AIKoClaw Helm repo
helm repo add aikoclaw https://charts.aikoclaw.io
helm repo update

# Install with production values
helm upgrade --install aikoclaw aikoclaw/aikoclaw \
  --namespace aikoclaw \
  --create-namespace \
  --set telegram.botToken=YOUR_BOT_TOKEN \
  --set domain=your-domain.com \
  --set telemetry.enabled=true \
  --set resources.cpu=8 \
  --set resources.memory=16Gi

# Or deploy via ArgoCD
argocd app create aikoclaw \
  -f https://aikoclaw.io/argocd/app.yaml \
  --sync-policy automated

🚀

Minimum

4 CPU cores
8 GB RAM
100 GB SSD
Public IP + SSL

Recommended

⚡

Production

8 CPU cores
16 GB RAM
500 GB SSD
Load Balancer + CDN

🌐

Enterprise

16+ CPU cores
32+ GB RAM
1 TB NVMe
Multi-region HA

Your FirstAI Employee