Workflow builder
A workflow builder that runs on your project board
A workflow builder is the tool you use to define how work moves from one step to the next. Most of them give you a drag-and-drop canvas of nodes and arrows. Agiflow takes a different shape: the workflow lives on a project board as tasks, statuses, work units, and workflow locks, and your AI assistant reads and updates that state through approved MCP tools.
Both shapes are valid. A canvas is better when the path is fixed and known in advance. A board is better when the sequence gets decided while the work happens, which is what most agent work looks like in practice.
Free forever for solo builders. No credit card required.
How a board-based workflow builder works
There is no canvas to draw. The structure of the board is the workflow, and four pieces carry it.
Tasks and statuses carry the sequence
Each unit of work is a task that moves through a status workflow. The order of work is whatever the board says it is right now, so changing the plan means moving a task rather than redrawing a diagram.
Work units group related tasks
A work unit bundles tasks that belong to one coherent chunk of work. It gives an assistant a scoped set of related tasks to reason about instead of the entire backlog.
Workflow locks handle coordination
A lock stops two people or two assistants from acting on the same item at the same time. This is the piece a diagram canvas usually leaves to convention, and it starts to matter once more than one agent is working.
MCP access connects the assistant
ChatGPT, Claude, Cursor, Codex, and other MCP-compatible clients read and update the board through tools you approve. The board stays the source of truth between sessions.
Diagram builders compared with a board-plus-agent model
These are two different tools for two different jobs. The comparison below is about fit, not about one approach being better than the other.
| Dimension | Drag-and-drop diagram builder | Board plus agent model |
|---|---|---|
| How the process is defined | Nodes and connections drawn ahead of time on a canvas | Tasks, statuses, and work units that change as the work changes |
| Best fit | Deterministic, repeatable paths where every branch is known | Open-ended work where scope and sequence are decided in progress |
| Handling a mid-flight change | Edit the diagram, then re-run or redeploy the flow | Add or move a task, and the next session reads the current state |
| Human review | Usually a dedicated approval node in the graph | A status the task sits in until a person moves it forward |
| Who executes the step | The platform runs the node | Your own assistant acts through approved MCP tools, because Agiflow does not run agents |
Choose a diagram builder when
- The steps are the same on every run, and you want them to be.
- Every branch and failure path can be enumerated before the first run.
- You want the platform itself to execute each step on a trigger or a schedule.
Choose a board when
- The next step depends on what the previous step actually produced.
- Scope changes mid-stream, and new tasks appear that nobody planned.
- A human needs to see and approve state before the work continues.
- More than one assistant or teammate touches the same project.
Bottom line: if you need a fixed pipeline that the platform executes for you, a diagram builder is the right tool and Agiflow is not it. If you need durable project state that an external assistant can read and update while the plan is still moving, the board is the better shape.
Workflow builder FAQ
Is Agiflow a drag-and-drop workflow builder?
No. Agiflow does not ship a diagram canvas where you drag nodes and connect them with arrows. The workflow is expressed as project structure: tasks move through statuses, work units group related tasks, and workflow locks control who or what can act on an item. External AI assistants read and update that structure through approved MCP tools.
When should I use a diagram-style workflow builder instead?
Use a diagram builder when the process is deterministic and repeats the same way every time, such as moving a record between systems on a schedule or routing a form submission. A canvas is a good fit when every branch is known in advance and the value is in automating a fixed path.
When does a board-based agent workflow fit better?
A board fits better when the work is open-ended and the sequence is decided as the work happens. Agent work usually looks like this: scope shifts, tasks get added mid-stream, and a human needs to review before the next step. Tasks and statuses hold that state without forcing you to redraw a diagram every time the plan changes.
Does Agiflow run the AI agents in the workflow?
No. Agiflow does not run, host, or execute AI agents and does not ship its own AI chat. You use the assistant you already have, such as ChatGPT, Claude, Cursor, or Codex. Agiflow provides the project board, the MCP tools, and the scoped access those assistants work through.
Keep reading
How workflows, work units, and locks fit together on an MCP-connected project board.
Build the workflow where the work already lives
Start free with 2 seats and 3 projects. Connect the assistant you already use, and keep tasks, statuses, and locks on a board that is still there in the next conversation.