Comparison
AI4Kanban vs.
Hermes Agent Kanban
Compare AI4Kanban and Hermes Kanban on specialist workflows, draft review, chat-app control, parallel tasks and follow-up work.
Specialist agents and workflows, ready to use
Built-in specialist agents and workflows cover product development and marketing content. Start with the workflow for your task instead of designing the team yourself. You can also create custom agents and workflows.
Hermes KanbanWorkers are Hermes profiles you set up with a model and skills. No built-in specialist workflows for UI design, copywriting or content production.
Worker profile
Name
Model
Skills
Both support
Task breakdown and dependencies
- AI4Kanban splits work into cards and subtasks, with dependencies controlling execution order.
- Hermes Kanban breaks a one-line task into child tasks and runs a child once its parents are done.
Parallel runs in git worktrees
- AI4Kanban runs independent cards side by side, each in its own git worktree.
- Hermes Kanban runs tasks in parallel, with a git worktree per task.
Review key drafts before implementation
Parts that need your judgment get a draft first. Review and edit it on the card; agents build from the approved version and handle the remaining details within the agreed scope. Catch changes here to reduce rework later.
Hermes KanbanTasks start from a text spec. Its docs describe no preview and approval of key drafts before execution.
Task spec
Goal
Approach
Acceptance criteria
Both support
A written spec
- AI4Kanban cards hold the scope and build steps.
- Hermes Kanban can rewrite a task into a goal, approach and acceptance criteria.
Feedback on the task
- AI4Kanban takes your changes in the card chat and updates the plan.
- Hermes Kanban takes your notes to the worker in task comments.
Settle requirements, then let agents execute
Approve key drafts to reduce rework, then check the delivery against the agreed requirements.
Plan the work with agents and settle the requirements and key drafts that need your decision. Agents then handle execution details, and independent tasks can run in parallel. You return for decisions and delivery checks without watching every session as it runs.
Hermes KanbanLight planning, fast execution: work starts as soon as it is broken down, the bar is adjusted along the way, and fixes are made in the worktree. That is a valid way to work, but it relies on you checking in as it runs, which limits how much work you can ship in a day.
Running task
Start
Adjust the bar
Fix in worktree
Both support
Agents that learn as they work
- AI4Kanban’s specialist agents note the drafts you send back or overrule, and finished cards are reviewed for decisions and preferences.
- Each Hermes profile keeps memory notes and writes its own skills from what it learns, including your corrections.
Task history
- AI4Kanban keeps plans, conversations and run records on the card.
- Hermes Kanban keeps a comment thread and run history on the task.
Compare the details
A marks the stronger side on each row.
Built-in specialist agents and workflows for development and content. Create your own. All built-in and custom workflows are free.
General-purpose Hermes profiles you set up. No built-in specialist workflows for UI design, copywriting or content production.
Planning settles what it can and asks you what is still open. Nothing is built until you start it.
A model breaks the task into a task graph without asking you; child tasks start on their own unless you turn that off.
Image, diagram, HTML/TSX, diff and storyboard drafts. Approved content becomes part of the execution requirements.
A text spec with goal, approach and acceptance criteria. No draft preview before execution in its docs.
Asked while planning or mid-build, each with options and a recommended answer; only dependent work waits, and it continues once you answer.
A worker pauses the whole task with a written reason; you comment, unblock it, and the worker starts again.
Each specialist agent notes the drafts you send back or overrule; finished cards are reviewed for decisions and preferences.
Each profile keeps memory notes and writes its own skills from what it learns, including your corrections.
Agents review what shipped and suggest follow-up work with reasons; a QA agent tests recent changes daily. Suggestions wait in triage for your decision.
Workers create child tasks to split up work in progress. Follow-up after delivery is a new task you create.
Each finished build is rebased and merged in turn; an agent resolves conflicts.
Worktrees are kept after the task. Merging back is not documented; conflicts go to a separate reconciliation task.
Scheduled agents run on a cadence you set.
One-off scheduled starts. Recurring work needs your own cron job.
Claude Code, Codex, Cursor, OpenCode and other coding agents run the work directly, on your own subscriptions; choose one per agent.
Workers are Hermes agents; a bundled skill lets one call Claude Code or Codex from the terminal.
A desktop app for cards, drafts, conversations and run status.
A CLI, a web dashboard and a Desktop app plugin.
Claude Code or Codex run tests and check the requirements as they build; AI4Kanban adds no second review pass, to avoid over-testing.
A reviewer profile checks each acceptance criterion and runs tests, sending work back until it passes.
In-app notifications work without Cloud; Slack and Lark alerts and remote control need Cloud, which is in invite-only preview.
Manage the board with /kanban from Telegram, Discord, Slack, WhatsApp, Signal and more, with task notifications.
Provider errors are retried automatically. A stopped run waits for you to resume it.
Heartbeats reclaim stalled tasks, and a task that keeps failing is put on hold.
A CLI that coding agents call. No public API.
A REST and WebSocket API, plus plugin hooks for task events.
Which should you choose?
Choose AI4Kanban if you
- Want ready-to-use specialist agents and workflows for coding, blogs and videos.
- Want to change key details in a draft before agents complete the work.
- Want agents to reuse your preferences so later tasks need fewer corrections.
Choose Hermes Kanban if you
- Already run Hermes Agent and want the board inside it.
- Want to manage tasks from Telegram, Slack, Discord or other chat apps.
- Want automatic recovery of stalled tasks and an API to build on.
Choose AI4Kanban for ready-to-use specialist workflows and key drafts that reduce rework; choose Hermes Kanban for chat-app control, automatic recovery and an API.
Compared against the Hermes Agent v0.21.6 documentation, checked October 2026.