What is Kaizen?
Kaizen is your organization's knowledge and work assistant, hosted on Kamiwaza. The information you need is usually scattered across product docs, wikis, code, chat, tickets, uploads, and other connected systems. Kaizen brings it into one permission-aware workspace with fast Search and a grounded, cite-everything Chat assistant.
Everything Kaizen shows you respects your organization's permissions. You only ever see sources you are already allowed to see — Kaizen never widens access.
Choose the Member, Administrator, or Developer view to see how each role meets the same governed system.
Other views: Member · Administrator · Developer
Use Full screen when a diagram needs more room. Inside the diagram, use the node finder to locate a service or role. Select a node to inspect its semantic passport and trace its authored upstream or downstream reach. Use PATH to choose two nodes and play their directed route.
What you can do
- Search your organization's knowledge and get back ranked source cards in a fraction of a second — no waiting on a model.
- Chat with an assistant that retrieves the right sources, reasons over them, and answers with inline citations you can click through to the original.
- Upload documents and ask questions grounded in them — privately, just for you.
- Build agents in Agent Studio: your own assistants with a custom persona, tools, and knowledge scope. Test them live, then share them.
- Collaborate — invite named colleagues into a chat or a Project, and work in one shared room.
- Schedule work — turn a repeatable request into a Task that runs on a schedule without an open browser.
- Organize knowledge with collections and saved prompts.
What makes Kaizen different
Grounded, not guessed. Chat answers are built from retrieved sources and carry citations, so you can verify every claim. Grounding is enforced by the backend, not left to the model's good behavior.
Permission-aware by design. Access control is applied inside the retrieval query rather than filtered out of the results afterwards, so a result set cannot drift outside what you are allowed to read. See Permissions & ACL enforcement.
It can act — carefully. Beyond answering, Kaizen can run code in an isolated sandbox, produce runnable artifacts, manage tasks, and invoke approved platform capabilities — always behind guardrails, budgets, approvals, and an audit trail.
Extensible by members. You do not have to wait on an admin to get a specialized assistant. You can build one, test it live, and share it yourself.
One pipeline, every agent. Streaming, retries, durability, policy, budgets, and audit are implemented once in a shared runtime. Every agent — default, custom, delegated, or scheduled — inherits the same behavior, so a capability can never be quietly weaker in one corner of the product.
Who uses Kaizen
Kaizen has two roles, mapped from your Kamiwaza account:
- Members — everyone. Members search, chat, upload documents, and build collections, Projects, agents, and Tasks.
- Admins — configure the deployment: which models power chat and embeddings, which knowledge sources are ingested, credentials, and feature availability.
You sign in with your normal Kamiwaza identity. Kaizen takes that identity as given — it never infers who you are from a request body — and uses it to enforce what you can see and do. See Roles & responsibilities.
How this documentation is organized
These pages serve three readers, and every page is badged for the one it was written for.
| If you are… | Start with |
|---|---|
| Using Kaizen day to day | Your first search & chat, then Working in Kaizen |
| Evaluating or operating a deployment | Architecture at a glance, Security model, Deployment topology |
| Building on or extending Kaizen | Member API, Building a tool or package, Building a connector |
Use the audience filter at the top of the sidebar to narrow these pages to a single lane, and the search box above it to jump straight to a topic.
If a term is unfamiliar, Concepts & glossary defines every one Kaizen uses.