leomuf

AI Software Architect — AI skill for Claude Code

AI community

An open-source, local-first AI software architect for Codex, Claude Code, GitHub Copilot, Antigravity and other coding assistants.

How to install AI Software Architect

This entry records only its repository, not the path inside it, so there is no exact command to give. Open leomuf/ai-software-architect and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What AI Software Architect does

An open-source, local-first AI software architect for Codex, Claude Code, GitHub Copilot, Antigravity and other coding assistants. Uses modular Agent Skills and Python MCP tools to turn requirements into explicit architecture decisions, ADRs, coding plans, and conformance reviews

Alternatives in AI

WeKnora — Open-source LLM knowledge platform: turn raw documents into a queryable RAG, an autonomous reasoning agent, an 26.9k ★
Open Science — Open Science Desktop — local-first, model-agnostic AI research workbench for macOS, Windows & Linux 1.5k ★
Get Physics Done — The first open-source agentic AI physicist, by Physical Superintelligence PBC (PSI) 752 ★

README

AI Software Architect

![AI Software Architect logo](assets/AISoftwareArchitect.png)

AI Software Architect is an open-source, host-native architecture agent for coding assistants. It helps developers clarify architecture-significant requirements, compare credible options, approve and record decisions, prepare coding handoffs, and review implementation conformance.

Maintained by Leonardo Muffato at [AUTOSOFT Engineering](https://www.autosoft-engineering.de).

OpenAI Plugin Directory Status

**Current status: Ready for submission.** The validated AI Software Architect v0.2.2 candidate has been prepared, but submission and approval have not yet been confirmed. See the canonical [`OpenAI Plugin Directory status`](docs/OPENAI_SUBMISSION_STATUS.md) for the submitted package identity and future review or availability updates.

Why AI Software Architect?

AI coding assistants are excellent at generating working code quickly, but they can begin implementation before making the architecture explicit. Important questions about constraints, quality attributes, component boundaries, data ownership, viable alternatives, and long-term trade-offs may therefore remain hidden inside one conversation—or never be considered at all. The result can work today while becoming unnecessarily coupled, difficult to test, or expensive to evolve.

AI Software Architect introduces a repeatable **architecture-first workflow** before substantial code generation. It asks only questions that can change a material decision, compares credible options instead of forcing a favorite pattern, requests approval, and records the outcome as durable repository artifacts. The coding assistant can then implement from an approved architecture contract and later review the code for drift.

Unlike a one-off architecture prompt, the project supplies a reusable method, fo