AD4M

What is AD4M and what is it for?

AD4M is the retired public repository for AD4M, a framework for building applications in which each user keeps their own identity and data instead of depending on a central service. It recorded the design's three building blocks — agents, languages and perspectives — so that separately built apps could share the same data and work together.

This repository holds an earlier version of AD4M, the Agent-Centric Distributed Application Meta-ontology. Its name records its status: AD4M's core continued under a separately named Coasys repository, and this one is best read as a historical record rather than a place to run anything from. What it preserves is the original account of AD4M's purpose — a shared layer for applications whose users hold their own identity and data — along with the concepts those applications were built from.

AD4M's claim was that an application can be assembled from three kinds of thing rather than one monolithic codebase. Agents are people with their devices, identified by a decentralised identifier instead of a login. Languages are adapters that wrap whatever technology actually carries the data — Holochain, IPFS, linked data or an ordinary back-end — and define the expressions an agent can create. Perspectives are an individual agent's own store of links between those expressions, the surface on which it perceives and writes.

The README walks through a small example: an agent creates a text note as an expression through a language, adds a perspective of its own, and links the note into it. Publishing that perspective turns a private graph into a shared neighbourhood that other agents can join. The retired repository's README remains the fullest first-party account of the design, and the current AD4M repository carries the project forward.

From 3 public sources, checked 23 September 2026.