Skip to content
archcore logo archcore logo archcore

Plugin Overview

Plugin

The Archcore Plugin is a smart AI plugin powered by the Archcore CLI — it uses your repository’s context to guide code generation. It installs on top of the CLI and gives you skills, multi-step tracks, built-in agents, and guardrails, all working against the same .archcore/ directory the CLI manages.

When to use the plugin

Section titled “When to use the plugin”

Use the plugin when you want the highest-quality experience inside your coding agent:

  • You work in Claude Code or Cursor day-to-day.
  • You want to type /archcore:adr and get a guided ADR flow, instead of explaining the document type to the agent.
  • You want multi-document tracks (PRD → plan, ADR → rule → guide, ISO 29148 cascades) to execute end-to-end.
  • You want the agent to refuse direct writes to .archcore/ and go through MCP every time.

If you want a minimal, direct integration — say, in a headless CI agent or an editor the plugin doesn’t support yet — start with the CLI instead.

How it fits with the CLI

Section titled “How it fits with the CLI”

Keep this analogy in your head:

  • Archcore CLI — the compiler. Reads .archcore/, builds the context graph, exposes it over MCP.
  • Archcore Plugin — the runtime. Applies that context inside your AI agent — skills, guardrails, workflows.

The plugin does not ship its own MCP server. It reuses the one from archcore mcp. This avoids duplicate-server conflicts in repos that already register archcore in .mcp.json or via claude mcp add.

What ships in the box

Section titled “What ships in the box”
  • 32 Skills — 18 document types, 8 intent commands, 6 multi-step tracks.
  • 2 Agents — a universal assistant (archcore-assistant) and a read-only auditor (archcore-auditor).
  • Hooks — session-start context loading, MCP-only write enforcement, post-mutation validation, and cascade staleness detection.
  • Rules — baseline guardrails that apply to every .archcore/ file interaction.

Without vs. with the plugin

Section titled “Without vs. with the plugin”

Without the plugin, ask for “a new service” and the agent:

  • guesses the folder structure
  • ignores conventions it has never been told about
  • produces code that drifts from decisions buried in docs
  • rediscovers patterns you’ve already written down

With the plugin, the same ask and the agent:

  • places files where your architecture says they belong
  • follows rules defined in .archcore/
  • respects ADRs and existing specs
  • reuses patterns (cpat) you’ve already captured

Next steps

Section titled “Next steps”