Kanban Board Software: What to Look For

Good kanban board software lets you shape the board to the way work actually flows, limit how much is in progress, see who changed what, and get your data out again. Every kanban app draws columns and cards, so judge the parts you cannot see in a demo: whether WIP limits block or only warn, whether swimlanes are real rows or a filter, what automation is allowed to change, how roles are set, whether AI assistants can read and move cards, and what the history and export keep. Below are eight criteria, the questions to put to a vendor, and where fenbs stands, including what it does not have.

This page is about buying criteria. If you only want to know what the free plans include, see the sister guide, free kanban board tools. If kanban itself is new to you, start with what is kanban.

1. Columns that match your workflow

The columns are your process written down. Check whether you can name, add and reorder them, whether a column maps to a status that reports and automations understand, and whether there is a backlog outside the board so the board shows only work you have committed to. A tool with fixed columns is not wrong, but then your process has to fit its columns rather than the other way around.

2. WIP limits: enforced or advisory

The Kanban Guide↗ (opens in a new tab) says the people in a kanban system “must explicitly control the number of work items in a workflow from started to finished.”

Software helps with that in two very different ways. Some tools block a card from entering a full column. Others only warn: GitHub’s documentation↗ (opens in a new tab) says a column limit in GitHub Projects “does not restrict anyone from adding cards that would exceed the column’s limit,” and highlights the column instead. Neither is wrong, but you should know which you are buying. How to pick the number is in kanban WIP limits.

3. Swimlanes and grouping

Swimlanes split a board into horizontal rows, by person, team, priority or parent item. In Jira, Atlassian’s swimlane documentation↗ (opens in a new tab) lists swimlanes by queries, stories, assignees, epics or spaces, set by a board or space admin. Other tools offer a “group by” on any field, which looks similar but behaves like a view rather than a board rule. Check whether dragging a card between rows changes its field, and whether lanes can carry their own limits. When lanes help and when they hide work is covered in kanban swimlanes.

4. Automation, and what it may touch

Automation in kanban software usually means rules: when a pull request merges, move the card to Done; when a card is added, set a field; archive anything finished for a week. Useful questions are less about how many rules you get and more about control. Who can create a rule? Does a rule respect WIP limits, or can it push a card past one? Is every automated change recorded as the rule, not as the person who wrote it? Free and lower plans often cap automation by runs per month, so ask what happens to a rule halfway through a month when the allowance runs out.

5. Roles and permissions

A board that only your team touches can live with everyone as an editor. Once a client, a contractor or a reporter joins, you want roles: who may read, add, move, comment and change settings. Check whether roles are fixed or definable, whether guests count as paid seats, and whether a refusal tells the person what they lack. The principles are the same for people and assistants; roles and permissions for humans and AI agents goes into them.

6. AI assistant access over MCP

More teams now want an AI assistant such as Claude, ChatGPT, Cursor or Copilot to read the board, pick up a card and report back. The usual route is MCP, which the Model Context Protocol site↗ (opens in a new tab) describes as “an open-source standard for connecting AI applications to external systems.”

Vendors differ on whether they ship their own MCP server, how the assistant signs in, and what it may do once connected. Trello’s official server, for example, uses OAuth and, according to Atlassian↗ (opens in a new tab), the assistant cannot do anything beyond what you can do in Trello yourself. The detailed checklist for an MCP server is in MCP server for project management: what to look for; for buying a board, three questions are enough:

• Is the assistant its own member with its own role, or does it act as you?
• Can you limit what it may do, and revoke it without touching your own sign-in?
• Are its changes recorded under its own name, so you can tell them apart from a person’s?

7. History

Every kanban tool keeps some activity log. What matters is whether it records every move, edit and comment, whether it names who made each change (a person, a rule or an AI assistant), how long it is kept on your plan, and whether you can see it per card. Without that, “who moved this back to To Do?” has no answer. The case for a full trail is in an audit trail for AI agents.

8. Export

Ask how your data leaves before you put it in. A useful export includes every card with its fields, comments and history, in a format another tool can read, and is available on the plan you are on, not only on a higher one. Try it in the trial with real cards. A tool you cannot leave is one you will think twice about adopting fully.

Questions to ask before you buy kanban software

1. Can a WIP limit stop a card from entering a column, or does it only warn?
2. Can we add, rename and reorder columns, and do reports follow the change?
3. Are swimlanes rows on the board, or a grouping in one view?
4. What can automation change, who can write rules, and what happens when the monthly allowance runs out?
5. Can we define our own roles, and do guests or clients count as paid seats?
6. Is there an official MCP server, how does an AI assistant sign in, and does it get its own role and name in the history?
7. How long is history kept on our plan, and does it name the person, rule or assistant behind each change?
8. What does an export contain, in what format, and on which plan?
9. Is there a mobile kanban app, and can it do everything the web board can, or only view?

Run the trial with a week of real work rather than a demo project. Most limits show up on day three, not in the sales tour.

Where fenbs stands

fenbs is deliberately simple kanban software, and some criteria above it does not meet. The lanes are fixed at four: To Do, Next Up, In Progress and Completed. You cannot add or rename columns, there are no WIP limits and no swimlanes, and there are no automation rules, sprints, due dates or a settable assignee. If your process needs a review column, an enforced limit or rows per team, a tool with those features is the better choice.

What it does offer lines up with the other criteria:

• Roles are set per company, with Client and Reporter as suggested roles, and every refusal names the permission that was missing and the role held.
• AI assistants are members of the board over MCP at https://fenbs.ai/api/mcp, signing in with OAuth or with a hand-issued token that has a name, scopes and an optional expiry. Revoking one stops it at once. Clients that only speak stdio use npx -y fenbs-mcp. See the MCP docs.
• History records every change with who made it, so an AI assistant’s edits are signed with its own name.
• Tasks have a kind (feature, enhancement or bug), a priority from 1 to 10 and an optional size from XS to XL, and the Decisions and rules page holds the rules every connected assistant reads first.
• Copy as Markdown copies the whole board as text, lane by lane.