CLI Overview

CLI

The Archcore CLI builds structured, git-native context for AI coding agents. It reads .archcore/, validates documents, builds the relation graph, and exposes it to every supported agent through MCP and session hooks.

Use the CLI when you want a minimal, direct integration — or when you’re using an agent the plugin doesn’t support yet.

When to use the CLI directly

• Your coding agent is GitHub Copilot, Gemini CLI, OpenCode, Codex CLI, Roo Code, or Cline — the plugin does not ship for these hosts yet, but the CLI’s MCP integration does.
• You want the smallest possible install (one binary, eight MCP tools, done).
• You’re scripting Archcore in CI, pre-commit hooks, or other non-interactive contexts.
• You want to understand how the full system works before adding the plugin layer on top.

How the CLI fits

• archcore init — creates .archcore/, detects coding agents, writes MCP config and session hooks for each.
• archcore mcp — starts the MCP server that agents launch as a subprocess.
• archcore validate / archcore status / archcore doctor — check structure, frontmatter, and relation graph health.
• archcore hooks — install/run session-start hooks per agent.
• archcore config — read/write .archcore/settings.json.

Want the higher-level UX with skills, tracks, and built-in agents? The plugin layers on top of the CLI.

See the Plugin path →

Supported agents

The CLI integrates with eight coding agents out of the box. MCP works with all eight; session hooks work with four of them.

Agent	MCP	Hooks
Claude Code	Yes	Yes
Cursor	Yes	Yes
GitHub Copilot	Yes	Yes
Gemini CLI	Yes	Yes
OpenCode	Yes	—
Codex CLI	Yes	—
Roo Code	Yes	—
Cline	Manual	—

For the full matrix and per-agent config paths, see Agent integrations.

Next steps

• Install — curl installer (macOS/Linux/WSL) or PowerShell (Windows).
• Quick start — initialize a project and create your first document.
• Commands — full reference for every archcore subcommand.
• MCP server — how agents connect and what tools they get.