Free & open source v3.39.43

AI DevOps

Automating development and scaling teamwork — designed to work 24/7, so you don’t have to.

A smarter, safer, faster AI harness — for maximum token‑efficiency, with every skill you need to design & build, all managed for you.

Your open-source AI DevOps managed agents, and OpenCode optimisation plugin, for automating projects, subagents, and background workers.

Ask for anything you want in your own words and it plans the work, builds it in isolated Git worktrees, and verifies everything’s working and secure before publishing — across code, infrastructure, SEO, content, and 150+ services.

Fully guided setup and best-practice, with tried & tested recommendations of free and low-cost services. All experience levels can use AI DevOps — it’s a great teacher, you’ll learn all you need to know as you go.

  • 400+ GitHub stars
  • 25,700+ commits
  • 14,800+ pull requests
  • MIT licensed
  • macOS & Linux
See how it works View on GitHub Open Source
Works with:

How it works

From one request to verified, shipped work

You bring vision, taste, priorities, and approvals. AI DevOps runs delivery, freeing your time for the things only you can do.

  1. Ask

    Describe the outcome in your own words, or use a slash command such as /full-loop. No prompt engineering required.

  2. Plan

    It recalls relevant memory, checks for duplicate work, and routes the task to the right specialist agent with a clear brief.

  3. Build safely

    Changes happen in an isolated Git worktree, guarded by secret handling, prompt-injection defences, and linters, so parallel work never collides.

  4. Verify & ship

    It designs and manages your CI/CD workflows, runs checks, opens a pull request, answers review feedback, and merges only when quality gates pass — then keeps the lessons for next time.

Work with it interactively, or let the pulse supervisor dispatch background workers around the clock.

Get started in seconds...

  1. Open the Terminal app on your computer

  2. Copy/paste: bash <(curl -fsSL https://aidevops.sh/install) Enter

  3. When set up, type opencode Enter. Then type /models Enter, Ctrl-A, and set up your OpenCode Zen, OpenAI, or Anthropic access. OpenAI GPT‑6‑Sol+ or Claude Opus‑5.5+ is recommended for complex work, with GPT‑6‑Luna+ or Sonnet‑5+ for fast triage and routine workers. Ctrl-F to Favourite models. Enter, Esc, Esc.

Now, just ask OpenCode in your terminal what aidevops does, how to use /onboarding, and which agent or command fits your next task.

And, for a smarter experience, try Tabby and Zed.

24/7 Development

Built in public, shipping around the clock

AI DevOps is the same system we use and evolve every day to build and optimise websites, apps, marketing, businesses, and client services. Every issue, pull request, and commit below is public.

17main agents
2,250+sub agents
2,060+helper scripts
100+slash commands
150+integrations
25+MCP servers
Issues loading
-- opened -- closed
Pull requests loading
-- opened -- closed
Commits per day loading

Your Git platform becomes a parallel teamwork audit-trail.

Quite possibly the best Git workflow harness in the world right now

AI DevOps treats every issue, pull request, and comment as continual project memory — who, what, where, when, why, and how — and runs high-volume GitHub automation inside API limits. Don’t take our word for it: every receipt is built in public.

99.9% of issues closed 17,500+ closed · 17 open 13,400+ pull requests merged 90.1% of closed PRs merged 330+ workflow labels status · origin · tier · dispatch 7× AI hours per human hour 124B+ tokens processed Top 600 on commit-history.com #559 by total GitHub contributions · Oct 2026

  1. Who

    Origin labels and assignee claims record whether a person or a background worker owns each task.

  2. What

    Issue briefs set the outcome, target files, acceptance criteria, and verification before work starts.

  3. Where

    Each task gets its own worktree branch and pull request, linked back to its issue, so parallel work never collides.

  4. When

    Timestamped, nonce-stamped dispatch claims stop two workers picking up the same task.

  5. Why

    Reviews, decisions, and follow-ups stay in the thread, so the next agent or person inherits the reasoning instead of guessing.

  6. How

    Signature footers log the version, runtime, model, time, and tokens spent. Workers can SSH-sign commits, and maintainer approvals and release tags are cryptographically signed.

Scales inside GitHub API limits

The budget-aware pulse routes reads to REST when GraphQL runs low, defers optional work in reserve mode, pauses dispatch at a circuit-breaker floor, and pre-warms caches — so round-the-clock issue, PR, and comment automation stays within rate limits.

Self-aware and self-improving

Ask it what it can do, which agent fits, or what it recommends — it answers from its own agent library and recommendations guide. Problems it hits become tracked issues, get fixed in public, and ship back as improved capabilities.

AI DevOps Features & Capabilities

AI DevOps (aka: aidevops, aiops, ai ops) is built for the gap between “the model can probably do this” and “the work is actually done, verified, safe, and worth the cost.” It gives AI assistants context, routing, security, git hygiene, collaboration, observability, memory, and follow-through.

OPS

Autonomous Delivery

Pulse supervision, missions, runners, mailbox coordination, budget tracking, and parallel worktrees turn large goals into verified PRs.

SEC

Security & Quality

Secret handling, posture checks, supply-chain scans, prompt-injection guardrails, cross-model review, and CI-quality gates.

AI

Agents, Skills & Models

Primary agents, specialist subagents, slash commands, imported skills, MCP tooling, and model-account pools for OpenAI, Claude, Cursor, Google, and local models.

BIZ

Beyond Code

Infrastructure, hosting, DNS, SEO, marketing, reports, content, video, browser automation, business operations, and personal productivity routines.

Token and time efficiency signals

  • Helper-first automation: deterministic scripts collect facts, compress output, and reduce repeated AI reasoning.
  • TOON-ready context: documented 20-60% token reduction vs JSON/YAML for indexes and registries.
  • Worktree discipline: isolated branches, PRs, CI gates, and review loops keep autonomous work auditable.
  • Domain routing: specialist agents avoid one-size-fits-all prompts for SEO, ops, security, content, finance, and more.

.agents file browser

loading
.agents/ Open on GitHub 
.agents/

▸ scripts/     helper automation that saves repeated AI reasoning
▸ workflows/   repeatable playbooks for delivery, git, reviews, and ops
▸ reference/   deeper guidance loaded only when needed
▸ tools/       service, browser, security, hosting, and platform docs

Loading the live GitHub tree...

AI DevOps Services & Integrations

AIDevOps connects your agents to the AI models, infrastructure, developer platforms, business systems, communications channels, creative apps, and automation services that make delivery happen. Names link to each provider; the matching agent guide loads into your AI only when it needs it.

AI Models & Compute

Hosting

Storage & Networking

Domain & DNS

Dev, Git & Agents

Websites, Apps & SEO

Mobile Apps

Design, UI & Creative

3D, Audio & Video Apps

Business & Payments

Email

Outreach

Security

Code Quality & Monitoring

Context Optimisation

Documents & OCR

Analytics & Performance

Browser & Automation

Team Chat

Private Messaging

Voice & Telephony

Missing something? Every integration is a plain Markdown agent guide. Browse the services and tools on GitHub, or ask aidevops to add a new one.

AI DevOps FAQ

Is AI DevOps really free?

Yes — AI DevOps is free of charge, and everything we recommend using with it is either free or the lowest-cost option we can find.

There is nothing to pay us, and the source code is included, transparent, and free too.

That openness is the best way we know to get it into the hands of as many creators and developers as possible.

OpenCode includes some free AI models, although eventually you will probably want a paid subscription with an AI provider.

Right now, we think a ChatGPT Pro or Claude subscription, with OpenAI GPT-6-Sol+ or Claude Opus-5.5+, is very good value.

Is AI DevOps open-source?

Yes — AI DevOps is open-source because AI tools and harnesses need high-level access to systems, tools, and computers.

Transparency and community collaboration are the safest way to keep evolving it as those capabilities increase.

What can I do with AI DevOps?

Anything that AI can do, but faster, cheaper, cleaner, with free and open-source options preferred, and with technical debt avoided by design.

Create and manage websites, apps, content, marketing campaigns, automations, chat channels, brands, companies, teams, and a lot of your personal life too.

Why AI DevOps, and not all the other AI tools?

AI DevOps is the only AI harness we know of designed from the start for team collaboration, where everyone has the same capabilities and can work in parallel.

That teamwork comes from the powers we inherited from Linus Torvalds’ Git project-management filing system.

Git is the real super-power here, because it makes collaboration, review, conflict resolution, and history sane at scale.

Most other AI tools are designed for individual self-management, which can become confusing and expensive when AI generates large amounts of competing instructions on every ask.

AI DevOps is built for teamwork, whether that means you and your background workers on one machine, or a whole company with many computers and servers collaborating across projects.

How do I get started?

Run the setup command, and if you do not know which one to use, use the first one.

Then open a new tab in your Tabby terminal, type opencode, hit Enter, and follow the first-use onboarding.

If onboarding does not show, type /onboarding in the chat and hit Enter.

It should walk you through the setup you need based on the first few things you think you might want to do with it.

What if I get stuck?

AI DevOps is self-aware enough that you can ask what you can do, what to do next, or for ideas.

If you find a bug or problem, use /log-issue-aidevops from any chat and explain what you are seeing and what you expected.

That comes straight through to the main GitHub project repo, where we can sanity-check it, ask clarifying questions, and solve it through AI DevOps self-improvement routines.

Who did all this?

Marcus Quinn. I live in Jersey, The Channel Islands, off the coast of France — creating cool things for work, and for fun.

AI DevOps started as my personal solution to the teamwork gaps I kept finding in basic vanilla AI tools.

I have added some of the people I work with to the repo to collaborate, and many more from the online communities I hang out in are already using it for daily work.

Let me know how you get on with it, with a shoutout or DM on X.com/marcuswquinn.

Ready when you are

Get Started with AI DevOps Today

Free and open source. One command installs it, then /onboarding walks you through the rest.