Archcore — Stop re-explaining your repo to every AI agent

A Git-native context layer for AI coding agents. The project keeps architectural decisions (ADRs), team rules, how-to guides, named patterns, action plans, and product requirements in your repository so Claude Code, Cursor, GitHub Copilot, Gemini CLI, and other agents follow your real architecture instead of guessing.

Why instruction files like CLAUDE.md and AGENTS.md aren't enough

Single instruction files are flat, tool-specific, and weakly structured. They pile up without type, link, or history. As a codebase grows, the agent has no way to find which rule applies to which directory, which decision blocks a refactor, or which spec governs a contract. The replacement here is eighteen typed document categories, named relations between them (informs, blocks, refines, supersedes), and versioned history — the same artifacts senior engineers already write but that agents can read, write, and reason about during real work.

Two ways to use it

Install the plugin for the best day-to-day experience inside Claude Code and Cursor, with architecture-aware /archcore slash commands. Or use the CLI directly with MCP and session hooks to support custom agent flows, minimal setups, and the eight AI coding agents the CLI integrates with today: Claude Code, Cursor, Gemini CLI, GitHub Copilot, OpenCode, Codex CLI, Roo Code, and Cline.

How the context loop works

Session hooks inject the right documents at the start of every conversation. MCP exposes tools (list, get, create, update) so the agent can browse and edit .archcore/ documents during a session. Slash commands like /archcore:context, /archcore:decide, and /archcore:plan give you a clean way to load, capture, and break down work without breaking flow.

Local-first, open source

Everything lives in .archcore/ inside your repository, versioned alongside the code it documents. No servers, no databases, no accounts, no telemetry. The CLI is a single cross-platform binary; the MCP server runs locally as a child process; install scripts pull release binaries from GitHub Releases. See the documentation or read the privacy policy for details.