AI DevOps Framework – aidevops

What it is – A command‑line framework that lets you orchestrate AI‑powered agents to carry out software‑delivery tasks from a high‑level intent → verified outcome. It is shipped as the aidevops npm/Homebrew package and works with the OpenCode AI‑assistant platform.

Key ideas

• Agents + knowledge base – 17 main agents (2 250+ sub‑agents, 2 200+ helper scripts) that encapsulate domain expertise (code, infra, marketing, research, creative, etc.).
• Durable context – All plans, decisions, evidence and task state are stored in the repository (Git) so work survives chat sessions.
• Safe Git workflow – Work is done in isolated linked work‑trees, verified with exact‑head checks, then merged via authority‑aware PRs.
• Model routing & token efficiency – Provider‑neutral tiers (simple, standard, thinking) automatically select the cheapest capable model; stronger models are reserved for high‑impact decisions.
• Automation layer – Pulse supervises registered repos, dispatches head‑less workers for issue‑scoped jobs, runs missions and routines on a schedule, and recovers from failures.
• Security & privacy – Secrets are stored with aidevops secret, prompt‑injection defence, supply‑chain checks, and audit trails.

Typical workflow

1. Install (brew install marcusquinn/tap/aidevops or npm i -g aidevops).
2. Connect model accounts (OpenCode, Anthropic, OpenAI, etc.) via aidevops model‑accounts‑pool.
3. Register your code under ~/Git, run aidevops init to add conventions.
4. Enable automation – aidevops auto-update enable, aidevops pulse …, aidevops init‑routines.
5. In an OpenCode session use commands like:
   • /full-loop "Implement feature X" – end‑to‑end delivery.
   • /skills recommend "TASK" – discover which agent can help.
   • /pulse – monitor and dispatch workers.
6. Review results with aidevops status, aidevops security, or generated reports.

Who it’s for

• Individual developers or small teams that want AI‑assisted CI/CD without writing custom scripts.
• Organizations looking to embed LLM‑driven agents into their existing Git‑based workflow while keeping security and auditability.

What you get out of the box

• CLI (aidevops) with > 100 slash‑style commands.
• Pre‑bundled agents for code generation, CI gating, infra provisioning, content creation, analytics, etc.
• Integration with OpenCode (the AI runtime) and optional Tabby IDE sync.
• Built‑in metrics, quality‑gate badges, and SonarCloud / Codacy reporting.

Limitations

• Requires the OpenCode platform (or compatible runtimes) and API keys for each model provider.
• Automation (Pulse/Workers) runs only on repositories you explicitly mark as "pulse": true.
• Some capabilities need local tooling (Docker, Firecracker, Tabby) to be installed.

Bottom line – aidevops is a full‑stack, AI‑first DevOps framework that turns natural‑language intents into reproducible, audited software delivery pipelines, leveraging a large library of specialized agents while keeping the work traceable in Git.