A kanban board for software developers, without the bloat

Four lanes, three kinds of task, refs that never change, and an MCP endpoint so your coding assistant can work the board too.

6 min read

If you want a kanban board for software work that stays out of your way, look for four things: a fixed, small set of columns; a task type that tells a bug from new work at a glance; a reference you can safely quote in a commit message; and a way for your coding assistant to read and update the board itself. fenbs is built around exactly those, and it leaves out the rest on purpose: no epics, no sprints, no story points, no custom fields, no automations.

What bloat actually costs a developer

Bloat is not only a crowded screen. It is the time between noticing a problem and having it written down. If filing a bug means choosing a project, an issue type, a component, a sprint and an estimate, you will not file it while you are in the middle of something else. You will tell yourself you will do it later, and some of the time you will.

It is also the cost of a board that means different things to different people. When columns can be added freely, every team grows a “Blocked”, a “Review”, a “QA” and a “Ready for release”, and half the cards in them are stale. A board with fewer places to put a card has fewer places for a card to be forgotten.

Three kinds of task

On fenbs every task is one of three kinds, and the product is named after them: Features, Enhancements and Bugs.

  • Feature. Something new that does not exist yet.
  • Enhancement. A change to something that already exists and works.
  • Bug. Something that is broken and needs correcting.

There is no fourth kind. No “chore”, no “spike”, no “story”. In practice almost everything a developer does is adding, revising or correcting, and the few things that are not still fit closely enough that a fourth bucket would cost more than it saves.

The kind is what colour means on the board. The design borrows the draughtsman’s pencils: blue adds, green revises, red corrects. Colour is kind, never status. A board with a lot of red on it reads as “a lot of bugs”, not “a lot of urgency”. Where a task is in its life is shown by which lane it sits in, and by nothing else.

Four lanes, fixed

The lanes are To Do, Next Up, In Progress and Completed. You drag a card between them, or use the arrows on the card. You cannot add a lane or rename one.

That sounds restrictive until you notice what Next Up does. It separates “everything we might do” from “what we have decided to do next”, which is the one distinction a backlog needs. It is also what makes the sentence “take the next thing in Next Up” meaningful to an AI assistant.

Each lane has its own sort, because lanes are read for different reasons: Custom, Urgent, Newest, Oldest, Changed, A–Z or Number. Custom is the hand-made order. Drop a card onto another card and it goes directly above it, and everybody on the board sees that order. Priority runs from 1 to 10, where 1 is the most urgent.

Refs you can put in a commit message

Every task has a ref such as FE-014, EN-015 or BU-016. The letters come from the kind and the number identifies the task. Two properties make these safe to quote.

  • Numbers are unique per board and never reused, even after a delete. A ref in last year’s commit still points at the same task.
  • The letters are worked out when the ref is shown, not stored. Change a task from a bug to a feature and BU-014 becomes FE-014 at once. Only the digits matter for lookup, so a link, a search or a tool call for the old ref still finds it.

If your work already has its own numbering, such as a ticket number from a client’s system, there is a separate box on every task called Your reference. Nothing parses it, and search covers it.

Your coding assistant works the same board

fenbs is an MCP server, so an assistant such as Claude Code connects to it directly. The home page describes Claude, ChatGPT and Cursor working the same board at the same time. There is one copy of the tasks, so there is nothing to map or sync.

Connect Claude Code
claude mcp add --transport http fenbs https://fenbs.ai/api/mcp
Then just ask
> what is on the fenbs.ai board?

  fenbs_whoami       → Editor on fenbs.ai
  fenbs_list_items   → 4 tasks

  In Progress  WEB-008  Board drag-and-drop between lanes        P1
  Next Up      WEB-006  MCP returns the missing permission       P1
  To Do        WEB-004  Dark mode tokens leak on the drawer      P2
  Completed    WEB-002  Refs must never be reused after a delete P2

The assistant connects with a token you issue under Settings, in Connect an AI assistant. The token is shown once, carries read, write and comment scopes, and can never do more than you can on the board. Every change it makes is recorded under its name, so when you read the history on Monday you can tell what you did from what Claude did. The connection guide lists the eight tools.

There is also AI context: notes on the board that every connected assistant reads before it touches your work, such as what you are building, how you like tasks written, and what must never be done. Notes can apply to everything or to one project. Change assistants and the new one reads all of it on arrival.

About the REST API

The API page documents the endpoints, and says plainly that API keys are not available yet. Until they are, MCP is the way to connect, and it covers the same ground.

Getting an existing backlog in

Most developers already have a backlog in a markdown file. Add many, next to + New task, takes it as it is. A heading above a list is treated as a section label. Indented lines under an entry become that task’s note. A line that starts with bug:, feature: or enhancement: sets that line’s kind, and lines without a marker take the kind you chose for the batch.

Fix the login redirect
bug: Payment webhook drops the currency
feature: Export the board as CSV

Everything is previewed before anything is created. Rows are created one at a time so the refs stay sequential, and if something fails partway, you are told which rows landed.

What you will not find

No epics, sprints, story points, custom fields or automations, and no configurable workflow. If your team plans in sprints and reports velocity, a tool designed around that will serve you better, and there are several good ones. fenbs is for the developer or small team who wants the board to take seconds, not meetings. How it works shows all of it, and plans start at $4.99 a month on the pricing page.

Questions people ask.

What task types does fenbs have?

Three: Feature, Enhancement and Bug. That is where the name fenbs comes from. There is no fourth kind.

Can I customise the kanban columns in fenbs?

No. There are four fixed lanes: To Do, Next Up, In Progress and Completed. Each lane has its own sort, and you can group tasks with optional projects and categories.

Are fenbs refs safe to use in commit messages?

Yes. Numbers are unique per board and never reused, even after a delete. If a task’s kind changes, the letters change but lookups match on the digits, so an old ref still finds the task.

Can Claude Code or Cursor update the board directly?

Yes. fenbs is a standard MCP server. An assistant connects with a token you issue, limited by read, write and comment scopes and by your own role, and its changes are recorded under its name.

Does fenbs have a REST API?

The endpoints are documented, but API keys are not available yet. Until they arrive, connecting over MCP needs no API key and covers the same ground.

Start with one thing.

There is nothing to set up first. Write one line and you’ve started.