AIWG now works with 11 AI coding tools — what's new in 2026.6.0

The short version: AIWG now sets up 11 different AI coding assistants, it's a lot easier to get AI agents to team up on a task, and your research papers are searchable instead of just sitting in a folder.

If you want to jump in:

npm install -g aiwg     # first time
aiwg refresh            # already have it? update + re-apply

New here? What AIWG actually does

If this is your first AIWG post, here's the one-paragraph version. Most AI coding tools — Claude Code, Cursor, Copilot, and so on — let you customize how the AI behaves: its instructions, its skills, the rules it follows. The catch is that every tool stores that setup in its own place and its own format, so if you use more than one tool you end up redoing the same work over and over. AIWG lets you write that setup once and then "deploys" it into whichever tools you use, in the format each one expects. One source of truth, many tools. That's the core idea; everything below builds on it.

1. One more tool: OpenHuman (that's 11)

OpenHuman is a free, open-source app for running your own personal AI. As of this release, AIWG can set it up too:

aiwg use sdlc --provider openhuman

You don't need to know its internals — AIWG puts everything where OpenHuman expects it, and we tested it against the real app to make sure the AI actually picks up your setup when it starts.

One honest heads-up: OpenHuman is still in beta, and there's a panel in its interface ("Installed skills") that currently looks in the wrong folder, so your AIWG skills won't appear in that list yet. They still load and work — it's a display bug on OpenHuman's side, and it affects their own installs too, not just AIWG's.

That brings the full list to: Claude Code, Codex, GitHub Copilot, Cursor, Warp, Factory, OpenCode, Windsurf, OpenClaw, Hermes, and OpenHuman.

2. Getting AI agents to work together — without the wiring

Two ideas here, and both are simpler than they sound.

Flows. A "flow" is just a recipe. Before, a multi-step job (like cutting a software release) was a long page of instructions you hoped the AI followed in order. Now you can write the steps as a short, structured checklist, and AIWG runs them one at a time — with safety checkpoints between steps so nothing important gets skipped. The same recipe adapts to different projects by changing a small settings file instead of rewriting the whole thing. This release even uses a flow to publish itself.

Missions. A mission is when more than one AI assistant teams up on a single job — and they don't all have to be the same tool. For example, one assistant can hand parts of the work to helpers running in a different tool, and a "conductor" keeps them coordinated. If you've ever wished you could split a big task across a couple of AIs and have them stay on the same page, that's the idea.

3. Your research papers, now searchable

If you keep a collection of papers or PDFs, this release turns it from a folder you dig through into something you can actually ask questions of:

• Search by meaning, not just keywords — find papers that are about a topic even if they don't use your exact words, and spot duplicates.
• See how papers connect — it maps which papers cite or relate to each other.
• Catch problems — it flags papers with quality or integrity issues and fixes up missing details.
• Read image-only PDFs — it can pull the text out of scanned documents that are really just pictures of pages.

Other nice touches

• Ask in plain English. The "find me the right tool" search now understands normal questions ("help me pick a framework"), not just exact keywords.
• Smoother across tools. A round of polish landed for Cursor, OpenClaw, Warp, and others so each tool gets cleaner, smaller setup files.
• Links point to the right place. Messages from the `aiwg` command now point at the public project page instead of an internal build address.

Want the full, detailed changelog? It's here. To try the release:

npm install -g aiwg

---

Tools & transparency

AIWG is open about how its content is made.

• Words: mostly AI-generated, then fact-checked and edited by a human.
• Facts and claims: every feature, command, version, and limitation here was checked against the actual 2026.6.0 release — nothing model-invented ships unverified.
• Hero image: AI-generated with ChatGPT (DALL·E) from a brand-specified prompt; no text or logos are AI-rendered.
• Final pass: human.

Net: AI-drafted, human-fact-checked.