Skip to main content

What is Alfred?

Alfred watches your inbox, processes raw inputs (meeting notes, conversations, documents), and produces structured Markdown records in an Obsidian vault. It maintains relationships between records using wikilinks, enforces schemas, and continuously improves vault quality through automated scans and knowledge extraction. Each Alfred instance runs on a dedicated server, accessible exclusively through your Tailscale mesh network. Everything — from dropping files into the inbox to managing Temporal workflows — is available through the REST API documented here.

How it works

1

Drop content into the inbox

Upload meeting notes, conversations, or documents via the API or directly into the vault’s inbox/ directory.
2

Curator processes the input

The Curator worker watches the inbox, invokes an AI agent to parse the raw content, and creates structured vault records with proper frontmatter, types, and wikilinks.
3

Janitor maintains vault health

The Janitor periodically scans the vault for broken links, invalid frontmatter, orphaned records, and other structural issues — then fixes them automatically.
4

Distiller extracts knowledge

The Distiller analyzes operational records and extracts latent knowledge: assumptions, decisions, constraints, contradictions, and syntheses.

Four tools, one vault

ToolPurposeTrigger
CuratorProcesses raw inbox files into typed vault recordsWatches inbox/ for new files
JanitorScans for and fixes structural issues (broken links, bad frontmatter)Periodic sweeps
DistillerExtracts assumptions, decisions, constraints from recordsOn-demand or scheduled
SurveyorEmbeds, clusters, labels, and discovers relationshipsPipeline execution

What you can do with the API

Next steps