-
Notifications
You must be signed in to change notification settings - Fork 26
Home
ssd edited this page Feb 24, 2026
·
2 revisions
Personal agentic infrastructure. Self-hosted. Always on.
Alfred is a self-hosted agentic butler — a layered AI architecture that captures your data, maintains your knowledge, executes your workflows, and communicates on your behalf. It runs continuously on your own infrastructure, so your AI doesn't just respond when prompted — it observes, thinks, acts, and learns.
| Layer | What it is | Alfred components |
|---|---|---|
| Interface | Where you interact | Telegram, WhatsApp, Slack, iMessage, SMS, email, CLI, TUI |
| Agent | The AI that reasons | Claude Code, Zo Computer, OpenClaw — pluggable backends |
| Kinetic | The engine that executes | Temporal workflows — durable, scheduled, auditable |
| Semantic | The memory that persists | Obsidian vault — human-readable, agent-writable knowledge graph |
| Data | The pipelines that capture | Omi transcripts, meetings, emails, RSS — raw signal in, structured knowledge out |
| Infra | Where it all runs | Mac Mini, VPS, personal cloud — your data, your control |
pip install alfred-vault
alfred quickstart
alfred upDrop a file into your vault's inbox/ folder and watch the Curator process it into structured, interlinked records.
Getting Started
Architecture
- Six-Layer Architecture
- Semantic Layer — vault as knowledge graph
- Kinetic Layer — Temporal workflow engine
- Agent Backends — Claude Code, Zo, OpenClaw
Semantic Layer Workers
- Curator — inbox processing (4-stage pipeline)
- Janitor — vault quality (3-stage pipeline)
- Distiller — knowledge extraction (2-pass pipeline)
- Surveyor — semantic mapping (4-stage pipeline)
Reference
Getting Started
Architecture
Workers
Reference