Honest Limits
Gemba Flow is the wrong choice for some teams and some jobs. This page says so plainly, and points you to what to use instead. If a tool only ever tells you where it shines, it has not earned your trust yet — so here is where it does not.
Who it is for — and who it is not
Gemba Flow is built for a solo developer or a 2–5 person team already using Claude Code, who is spending more time inventing workflow ceremony than shipping. Six agents with non-overlapping authority give that team structure without a heavy process to learn.
If your team is larger than ~20 people, Gemba Flow is probably too lean. Look at BMAD-Method instead: its 21 specialized agents model a full agile organization and it is tool-agnostic. That depth is genuine overhead for a small team, but it is a real strength once you have an org-sized process to mirror. Picking the heavier tool is the right call when you actually have the headcount.
If you are an engineering lead or buyer who needs pricing, governance, security, and ROI material to make a rollout decision: that content is not here yet. It is planned for a later phase, and we would rather say “not yet” than pad this site with thin enterprise pages.
When to choose something else
- You want a full agile-org simulation. If you want product-owner, scrum-master, and analyst agents running a complete ceremony, that is BMAD-Method’s design, not ours. Gemba Flow deliberately stops at six roles.
- You are leaving Claude Code. Gemba Flow is coupled to Claude Code on purpose — it builds on Claude Code’s CLI, context window, and sub-agents. If you are moving to a different harness (Cursor, Codex, Cline), adopt that harness’s own workflow; Gemba Flow will not come with you.
- You need an in-browser interactive sandbox. There isn’t one. A static end-to-end walkthrough plus video covers the evaluation need today; a live sandbox is too costly to build well right now and may come in a later version.
- You need multi-language documentation. This site is English-only for v1.
- You need a full skill-authoring API reference. v1 documents using the built-in skills. Writing your own skills is a depth topic for a later phase.
In short: if your need is on this list, you will be happier with the tool we named than with a forced fit here.
Known limitations, today
These are real and current. None is hidden behind a “coming soon.”
- Install numbers are a proxy, not a count. v1 measures adoption by a click on the install command, not by confirmed installs. That over- or under-counts real installs. A more accurate, opt-in signal is planned for a later phase (see Telemetry below).
- The “agent authority” idea is no longer novel on its own. Claude Code’s native sub-agents now describe the same separation-of-duties principle. Gemba Flow’s actual edge is enforcement — branch protection, account separation, and a merge that only a human can perform — not the idea of boundaries itself. If you only want the idea, you already have it natively.
Telemetry and privacy
v1 ships zero telemetry. The framework does not phone home, and there is no hidden usage tracking baked into the tooling.
A later phase will add an opt-in install beacon: it is off unless you turn it on, it sends no personally identifiable information, and you will be able to read exactly what it sends in the route’s source before enabling it. Until that route ships, there is nothing to inspect because there is nothing being sent.
Still the right fit?
If none of the above ruled you out, you are exactly who Gemba Flow is for. The Quickstart takes you from install to your first agent-authored pull request, and Structured Handoffs explains how the six agents keep that pull request honest.