AH
🚀

Getting Started

Five steps to go from zero to a working AI memory system. Follow them in order — each builds on the last.

12345
1

Set Up Your Environment

15–20 minutes

Install dependencies and clone the repository

Get your development environment ready — clone the Memory Platform monorepo, install system dependencies (Node.js 20+, Python 3.11+, Docker), and run the bootstrap script.

Checklist

  • Clone the repository from GitHub
  • Install Node.js 20+, Python 3.11+, and Docker
  • Run make bootstrap to install all dependencies
  • Verify with make check — all green
Repository READMEBootstrap Guide
2

Configure Providers

10–15 minutes

Connect your AI model providers

Configure LLM provider API keys and set up the provider adapter. The platform supports OpenAI, Anthropic, Google Gemini, local models (Ollama, LM Studio), and custom endpoints through a unified interface.

Checklist

  • Obtain API keys for your chosen providers
  • Run hermes setup or edit .env with your keys
  • Configure the provider adapter in config/providers.yaml
  • Test connectivity with a simple query
Provider SetupProvider Adapter ADR
3

Understand Memory

20–30 minutes

Learn the memory layer fundamentals

Understand the three memory tiers — ephemeral (in-session), working (cross-session), and archival (long-term). Learn how Honcho manages memory with semantic search, compaction, and user-scoped isolation.

Checklist

  • Explore the three memory tier model
  • Start the Honcho server with docker compose up honcho
  • Run the memory demo: make demo-memory
  • Review the Honcho capability matrix
Memory Systems FacultyHoncho Setup
4

Build Your First Agent

30–45 minutes

Create a conversational agent with memory

Build a working conversational agent that remembers context across sessions. Use the context builder for prompt assembly, the policy engine for content filtering, and the event bus for inter-agent communication.

Checklist

  • Create a simple agent using the memory SDK
  • Configure the context builder with token budget
  • Add a policy for PII redaction
  • Test across multiple sessions — memory persists
Agent Design FacultyQuickstart Tutorial
5

Deploy & Monitor

20–30 minutes

Go to production with confidence

Containerize your agent, deploy with Docker Compose or Kubernetes, and set up observability. The platform includes distributed tracing, context usage monitoring, and structured logging out of the box.

Checklist

  • Containerize with Docker — use the provided Dockerfile
  • Deploy with docker compose up -d
  • Check the status dashboard at /status
  • Set up alerts for context overruns and provider errors
Status DashboardDeployment Faculty

You're ready to build

Completed all five steps? Dive deeper into specific topics at the AI University or check your system health on the dashboard.

Explore AI University →Check Dashboard