The ecosystem

Every project, and what it is the authority for.

Knowledge Islands is not one program. It is a set of repositories with deliberately separate authority — philosophy, engineering discipline, agentic capabilities, portable standards, command-line tools, MCP adapters, and the infrastructure that delivers them. This page names all of them, says what each one owns, and sends you to the owner.

Only public repositories are listed.

Principal knowledge bases

Where the thinking is kept.

Knowledge Islands Principal holds the ecosystem's reasoning: the philosophy, the conceptual model, and the engineering discipline that turns them into buildable practice. Everything else in this list is downstream of it.

Agentic capabilities

What an agent is actually given.

A capability is a typed, published thing an agent can be handed — a skill, an agent definition, an MCP server, a hook, an eval. These repositories define what each one contains and how it is delivered to a runtime.

Portable standards

The contracts implementations must keep.

Normative, portable contracts, written only once a concept is proven and implementations have supplied evidence. An Active specification then governs anything claiming conformance within its scope.

Command-line tools

What you can install today.

The four executables that are released and installable right now. Each has a stable installer endpoint on this site that redirects to the installer published at the exact version named on its card, so the page, the endpoint, and the advertised release always move together. Everything else in this directory is read, cloned, or consumed rather than installed.

MCP servers

Adapters onto the systems where work already lives.

Model Context Protocol servers that give an agent bounded access to a real system — a knowledge base, a mail account, a tree of repositories. Each one defaults to read-only, validates every path or identifier it is handed, and hides its destructive tools until an operator opts in. None of them is published to a package registry yet; each is built and registered from source.

Platform and delivery

How the rest of it reaches a machine.

Infrastructure and transport. These repositories run work, package releases, and publish pages; none of them is an authority on knowledge, capability, or standards.

Why the list is shaped this way

Separate authority is the point, not an accident of history.

The split exists so that a philosophical claim cannot quietly become an engineering mandate, a CLI implementation cannot quietly define a standard, and a page on this site cannot quietly become the source of truth for something it merely describes. Each project has a page here that says what it solves, who it is for, what state it is in and what it does not do; each repository remains the authority for its own interfaces and releases.

One registry stands behind this page, including the released tools. A tool is a project that happens to ship a binary, so it carries a version and an installer where the others do not; the site still hosts no artefact and owns no executable behaviour. Every promise about a specific release belongs to the repository that published it.