concepts

Contexts are the backbone. Retrieval is where it gets useful.

The home page should stay readable. This section is for the person who wants to know whether the architecture holds up: how contexts keep the graph permission-aware, and how apps or agents then find useful data, follow links and keep working across pods.

Context permissions

Contexts are the current permission boundary in the reference implementation. Read this first if you want the primitive the other concepts rely on.

Graph retrieval

Retrieval starts with resources and links, not document chunks.find gives a permission-aware foothold; traversal does the deeper work.

Linked Open Data and federated grants

WebIDs, app DIDs, aliases and grants are the identity side of the model: apps should connect to pods without every product becoming the account system.

MCP over pods

MCP is another client surface over the same pod rules. Multi-pod MCP is interesting because an agent can query independent data spaces instead of one central memory service.

Linked Data CRUD

CRUD sounds ordinary because the web is ordinary. That is the point: resources can be addressed, read and written over HTTP instead of being hidden behind one SDK.

Read context permissions Read graph retrieval Read federated grants