marcusquinn/aidevops

AI DevOps Framework – aidevops

What it is

• aidevops is an open‑source framework (MIT‑licensed) that lets you run AI‑driven automation across the whole software‑delivery lifecycle and many business domains. It ships as a CLI (aidevops) and as an OpenCode plugin, so the same commands can be invoked from a local shell or from the OpenCode AI‑assistant UI.

Core idea

• Instead of prompting a large language model with the entire task each time, aidevops breaks work into agents (14 main agents, >2 200 sub‑agents) that each have a narrow responsibility (e.g., build, SEO, security, marketing). A lightweight supervisor (“pulse”) routes work, tracks token usage, isolates secrets, and keeps Git history clean.

Key capabilities

Area	What the framework can do
Code & CI	Create work‑trees, open PRs, run linters, audit quality gates, roll back via audited bundles.
Infrastructure	Manage hosting, DNS, monitoring, self‑hosted runners, deployment copies, and secret vaults.
Product & Business	Draft PRDs, road‑maps, growth experiments, invoices, marketing copy, SEO reports.
Creative & Content	Generate UI mock‑ups, decks, videos, reports (Markdown → HTML/PDF) with evidence badges.
Safety & Governance	Context‑window compaction, cross‑model checks, vault‑encrypted secrets, audit logs, token‑budget reports.
Automation glue	2 010+ helper scripts, 100+ slash commands, and a set of “/” commands (e.g., /pulse, /optimize‑tiers, /report‑token‑use).

How you use it

1. Install globally with npm (npm install -g aidevops).
2. Run aidevops init inside any repository to bootstrap the framework.
3. Invoke high‑level commands (aidevops pulse, aidevops /design‑artifact, aidevops secret) or call a specific agent via its markdown module.
4. The supervisor decides which model tier to use (cheap model for routine work, GPT‑6‑Astra for high‑level reasoning, Claude as fallback) and keeps the work auditable.

Installation & runtime

• CLI: npm package (aidevops), also available via Homebrew (marcusquinn/tap).
• Runtime: Bash scripts + TypeScript (run on Bun). Works with OpenCode’s control‑plane and any OpenAI‑compatible API; recommended models are GPT‑5.6 Luna, GPT‑5.6 Terra, and GPT‑6 Astra.

Target audience

• Engineers who already use Git‑based DevOps and want AI to handle the “glue” (branch management, PR creation, CI monitoring) without writing prompts for every step.
• Product & marketing teams that need AI‑generated assets but still require version control and budget tracking.
• Organizations that need a reproducible, auditable AI‑automation layer (the framework records decisions, token spend, and secret handling).

Project health

• Actively maintained (latest version 3.32.323, CI badges show passing builds, SonarCloud quality gate, CodeFactor, Codacy). Over 2 000 helper scripts indicate a sizable code base.
• Documentation is extensive (design docs, agent catalog, security vault guide, reporting templates).
• Community contribution is encouraged; MIT license permits commercial reuse.

Why it matters aidevops tackles the “orchestration gap” that most LLM‑based tools have: they can write code, but they don’t manage the surrounding workflow (branch hygiene, secret safety, token budgeting). By embedding DevOps best practices into the AI agent layer, the framework aims to make AI‑augmented development scalable and cost‑effective.

---

All details above are taken directly from the repository’s README; no additional features have been inferred.