AKIOS Logo AKIOS
EN | FR

Quickstart Guide

Current Version: 1.6.0

Go from zero to your first secure AI workflow in 15 minutes. This guide walks you through everything step-by-step.

Current version: 1.6.0. 6 core agents, 7 LLM providers (including local via Ollama), parallel execution, plugin system, 44 PII patterns, Merkle audit proofs, HTML compliance reports, and ablation benchmarking.

What You'll Build

By the end of this guide, you'll have:

  • ✅ AKIOS installed and configured
  • ✅ Your first AI workflow running with real LLM calls
  • ✅ Understanding of project structure and how to create custom workflows
  • ✅ Knowledge of security features and audit logs

Prerequisites

  • 5 minutes of your time
  • An API key from OpenAI, Anthropic, Grok, Mistral, Gemini, or AWS Bedrock (get free keys below) — or use Ollama for free local models
  • Docker installed (for cross-platform) OR Python 3.9+ (for Linux)

Step 1: Install AKIOS (2 minutes)

Choose your platform:

Linux users:

pip install akios>=1.6.0

macOS/Windows users or teams:

curl -O https://raw.githubusercontent.com/akios-ai/akios/main/src/akios/cli/data/wrapper.sh
mv wrapper.sh akios
chmod +x akios

That's it! Installation done.

Step 2: Create Your Project (1 minute)

# Linux (pip)
akios init my-project

# macOS/Windows/Docker
./akios init my-project

# Go into your project
cd my-project

What just happened? AKIOS created a complete project structure:

  • templates/ - Pre-built workflow examples
  • workflows/ - Where you'll put custom workflows
  • data/input/ and data/output/ - For file operations
  • audit/ - Security and compliance logs
  • config.yaml - Runtime settings

Step 3: Configure Your API Key (2 minutes)

The setup wizard makes this easy:

# Linux (pip)
akios setup

# macOS/Windows/Docker
../akios setup

The wizard asks you:

  1. Which AI provider? (OpenAI, Anthropic, Grok, Mistral, Gemini, AWS Bedrock, or Ollama)
  2. What's your API key? (Paste it securely — or skip for Bedrock IAM / Ollama local)
  3. Set a budget limit? (Default $1.00 per workflow)

Get free API keys:

Step 4: Run Your First Workflow (5 minutes)

# Linux (pip)
akios run templates/hello-workflow.yml

# macOS/Windows/Docker
../akios run templates/hello-workflow.yml

What you'll see:

✓ Loading configuration...
✓ Step 1: LLM generates greeting
✓ Step 2: Saves to file
✓ Workflow completed
Cost: $0.0012 | Time: 3.2s

Check the results:

akios files output
cat data/output/run_*/hello-ai.txt

You just ran your first secure AI workflow! 🎉

Step 5: Explore Security Features

Check system status:

akios status

Shows security level, PII redaction, budget remaining, and audit status — now with a rich terminal UI.

Activate the security cage:

akios cage up
akios cage status

Scan files for PII before processing:

akios protect scan data/input/document.txt
akios protect preview templates/hello-workflow.yml

Run diagnostics:

akios doctor

Checks your system for compatibility issues (kernel, seccomp, cgroups, Docker).

View the execution timeline:

akios timeline

Get the latest output as deployable JSON (for CI/CD):

akios output latest

View audit logs:

akios logs --limit 10
akios audit verify

Every action is cryptographically logged for compliance.

See available templates:

akios templates list

Generate a compliance report:

akios compliance report templates/hello-workflow.yml

Create Your Own Workflow

Copy and customize a template:

cp templates/hello-workflow.yml workflows/my-greeting.yml
nano workflows/my-greeting.yml  # Edit the prompt
akios run workflows/my-greeting.yml

Common Tasks

Process a document:

echo "Contact john@example.com" > data/input/doc.txt
akios run templates/document_ingestion.yml

Clean old runs:

akios clean --old-runs

Troubleshooting

"API key not valid" → Run akios setup --force

"Permission denied" → Linux: sudo usermod -aG docker $USER then logout/login

"SIGTRAP crash with sudo" → Fixed in v1.0.5. Upgrade: sudo pip3 install --upgrade akios

"Workflow failed" → Check akios logs --limit 5 or see Debugging

What's Next?

You're ready to build secure AI workflows! 🚀

ESC