The platform does a lot. But the daily reality is repetitive, knowledge-heavy, and easy to lose track of.
The same scripts, the same lookups, the same documentation — rebuilt by hand, project after project.
The right JSP call, the right API for the right context — known by a few, written down by fewer.
Batch work, audits, and write-ups wait for a person with time. Nothing runs on its own.
What if every one of those tasks had a specialist who never forgets — and works while you sleep?
Kage-gumi is a team of AI operatives — each one given a single job, its own knowledge, and real reach into the systems we use every day.
Every operative is an expert in exactly one thing. No jack-of-all-trades. Depth over breadth.
They don't just talk — they reach into WebCenter, the browser, and Miro and change real state.
Tasks route to the right operative, run in the background, and report back. Some run overnight, unattended.
Knows the full WebCenter SDK: every JSP endpoint, every attribute, every error pattern. Ask in plain language, get working answers. Mei prepares the ground before anything else happens.
Built around WebCenter — Workflow, Rule Engine and Dashboard scripts, plus integration and migration specs and Workato recipes & diagrams. Each context has its own rules, and Ren knows them all. He doesn't stop at writing the code: he uploads scripts straight into WebCenter himself, no copy-paste step. Saved, versioned, diffable. Tempered by the work — refined until it holds.
Turns messy work into clean documentation, then pushes the work itself live. After Ren builds something, Haku writes it up — what it does, what it touches, how to use it — and deploys scripts straight into WebCenter via Playwright. Searchable, shareable, stored. The inspector's eye she first opened, 透視 Tōshi, has since grown into an operative of her own.
Runs scheduled tasks while I sleep. Surfaces his own proposals for the next run, queues scripts for review, and delivers a morning digest. Launch a proposal and it becomes a real, tracked dashboard task. Dragons don't rest. Neither does Ryū.
Sits in the shadows of a call and listens. Captures the audio, transcribes it on-device with faster-whisper — nothing ever leaves the machine — and hands back a clean summary: decisions, action items, open questions. A meeting stops being something you re-type from memory and becomes reusable text that feeds back into project work.
The ear hears; the ink keeps. After Chō summarises a meeting, Sumi files it into a durable, per-customer record — a clean dossier that outlives the call. Tag a meeting to a customer and it's archived for good. Faithful, factual, never invents an attendee or a date.
Where the others build and deploy, Tōshi reveals. She X-rays any WebCenter — dashboard topology, page inventory, the wiring and hidden structure nobody wrote down — and hands back a map of what's really there. Every build starts by seeing the ground first, so Tōshi is never idle. The shadow that reads the blueprint behind the glass.
The operatives aren't different AI models. They're the same model — shaped into specialists by what they're told and what they remember.
The CLI built the house. The LLM is the mind living inside it. MCP is the arms reaching into the world.
No copy-paste, no dead end at "here's some text." A task flows all the way through to a result that lands.
Through MCP and browser automation, the crew operates the same tools a person does — at machine speed and without forgetting a step.
Ryū, the Night Runner, takes the queue overnight — unattended, on a schedule, inside a strict token budget.
Kicks off on its own at a set hour. No one sitting in a chat window.
Respects a hard token cap per operative and per day — spend stays predictable.
Every run lands as a tracked task with a morning digest of exactly what happened.
Load the queue before you leave. Wake up to work that's already done.
Chō listens to a meeting and transcribes it locally — no audio, no transcript, nothing sent to a cloud service. Sumi then files the result into a durable record.
Speech-to-text runs on-device. Meeting audio never goes online. Out comes a structured summary: decisions, actions, open questions.
The summary is filed into a clean, faithful record that outlives the call. Factual by rule — never invents an attendee or a date.
Be present in the room — not the note-taker.
A request doesn't stop at code. It goes all the way to live, documented, and committed — with full traceability.
A general AI tool improvises from public docs. This one draws on a real, accumulated body of WebCenter craft — years of scripts, patterns, and solved edge cases.
When Mei answers or Ren builds, it's lifting from genuine, battle-tested solutions — not hallucinating from a manual.
Each context — Dashboard, Rule Engine, Workflow node, SDK — has its own rules. The crew knows which is which.
The AI didn't replace the experience. It made years of it instantly usable.
Where it's heading: from a personal assistant to a build platform — entire projects designed as work items, executed by operatives, and deployed in one motion.
Describe what to build. It's broken into executable tasks and reviewed before anything runs.
Each item is dispatched to an operative, built, and returned for review. Overnight, the whole queue runs itself.
Accepted work deploys into WebCenter, gets documented, and is committed. Full traceability.
A backlog, executed by the crew — without the overhead. Your build pipeline is Ren, your overnight CI is Ryū, and every result is documented and versioned.
A year of real, shipped capability — and a clear path forward. The crew grows deliberately, never all at once.
"You don't need to be into AI to find this useful. You just need WebCenter work that's taking more time than it should."