Archy
Let your coding agent scan a Python repo for architectural drift, layer violations, and structure smells before refactors or releases.
Overview
archy is a MCP server for the Ship phase that senses architectural structure and health signals in Python codebases for agent-assisted review.
What is this MCP server?
- stdio MCP server (archy v0.13.3) installable via uvx from PyPI
- Architectural sensor focused on Python codebases, not generic linters
- Surfaces structure and dependency signals agents can use in pre-merge review
- Runs locally through stdio transport for Claude Code, Cursor, and Codex
- Pairs with human code review when you lack a dedicated architecture doc
- Server version 0.13.3 on PyPI package identifier archy
- Transport type stdio with runtimeHint uvx
Community signal: 1 GitHub stars.
What problem does it solve?
Python repos grow messy module boundaries and hidden coupling faster than solo builders can manually audit before every merge.
Who is it for?
Indie Python backend or CLI authors who want lightweight architecture checks during agent-driven code review.
Skip if: Teams that need runtime APM, security pentesting, or first-class support for non-Python monorepos.
What do I get? / Deliverables
After you add archy, your agent can ground refactors and reviews in repeatable architectural observations instead of guesswork about package layout.
- Agent-readable architectural observations about Python repo structure
- Review inputs you can fold into refactors, ADRs, or merge checklists
Recommended MCP Servers
Journey fit
Canonical shelf is Ship because architecture sensing is most valuable when you are reviewing code quality and release readiness, even though it also helps during backend work and production iteration. Review is where structural feedback prevents shipping spaghetti modules and unclear boundaries in Python services.
How it compares
Python architectural sensor MCP, not a generic linter skill or cloud monitoring integration.
Common Questions / FAQ
Who is Archy for?
Archy is for solo builders and small teams maintaining Python code who want their AI coding agent to reason about structure and boundaries during review and refactor work.
When should I use Archy?
Use Archy before large refactors, when onboarding to an unfamiliar Python service, or during ship-phase review when you need evidence about module layering and architectural drift.
How do I add Archy to my agent?
Register the stdio MCP server io.github.hslee16/Archy with runtime hint uvx and the PyPI package identifier Archy, then restart Claude Code, Cursor, or your MCP-compatible client.