Skip to content

Workflow automation

Automate the work behind your apps.
With people in the loop where it matters.

Run multi-step processes on your data and APIs. Combine queries, code and AI steps, with human approval before actions that need a decision.

One run, start to finish
Run 4821Purchase requestQueued
  1. Not startedTriggerPurchase request submitted
  2. Not startedAI agentCheck the request against policy
  3. Not startedApprovalFinance approval
  1. The run starts when this comes into view.

The approval request appears here when the run reaches Finance.

Human in the loopNew

Add an approval step where a person should decide.

The run stops at that step, waits for a decision and continues on the path that matches it.

  • The run waits. Nothing after the approval step runs until someone decides.
  • A person decides. The approver sees the request with the data the workflow gathered.
  • Both answers have a path. Approved and rejected each continue their own way, with the decision on record.
Put a person before
  • Payments and refunds
  • Access changes
  • Anything an AI agent wants to write
  • Customer messages
  • Production data fixes
Build workflowsExplore ToolJet MCP

Build it on a canvas, or let your agent draft it.

Connect steps on a visual canvas. Each one reads or writes your data, runs code, makes a decision or waits for a person.

Or have your coding agent draft the workflow, through ToolJet MCP. It lands on the canvas as an editable draft.
Claude CodeCodexCursor
Triggers

From a ToolJet app

A button, a form submit or any event in an app starts the workflow and gets the result back.

From a webhook

Another system calls the workflow's URL when something happens on its side.

On a schedule

Run it every hour, every night or every Monday morning.

Run and monitorEnterprise security

Every run leaves a record.

Workflows run on your server, not in someone's browser. Open any run to see where it is and what each step did.

  • A log for every step

    Open any run to see each step's status and output, including who approved it.

  • Limits you set

    Timeout and memory limits keep one heavy run from affecting the rest.

  • Your permissions apply

    The same roles, encrypted credentials and audit logs as your ToolJet apps, on self-hosted ToolJet.

What teams automate

The processes teams automate first.

A person decides at this step.

Used by teams at

SwisscomDuolingoGojekOrangeNavitas

Frequently asked questions

What is ToolJet workflow automation?
ToolJet Workflows automate multi-step business processes on your own data and APIs. You connect steps on a visual canvas: queries to 100+ data sources, JavaScript, conditions, loops, AI agents and approval steps. A workflow runs on the server and can be started from a ToolJet app, a webhook or a schedule.
What is a human-in-the-loop workflow?
A human-in-the-loop workflow is an automated process that stops at a defined point and waits for a person to decide before it continues. In ToolJet you add an approval step where a decision is needed, for example before a payment, an access change or an action proposed by an AI agent. Everything before and after that step stays automated.
How does the approval step work in ToolJet workflows?
When a run reaches an approval step it pauses. The approver reviews the request with the data the workflow has gathered so far, then approves or rejects it. The run continues from that step on the path that matches the decision, and the decision is kept with the run's log.
How are ToolJet workflows triggered?
A workflow can start in three ways: from an action in a ToolJet app, such as a button or a form submit, from a webhook called by another system, or on a schedule.
Can I use AI agents inside a workflow?
Yes. An AI agent is a step you add to a workflow. You give it instructions and the tools it is allowed to use, and it returns a result the next steps can act on. Pair it with an approval step when a person should review what the agent proposes before the workflow writes any data.
Can Claude Code or Codex build ToolJet workflows?
Yes. Through ToolJet MCP, coding agents such as Claude Code and Codex can discover your data sources, then create, edit and validate workflow drafts. The result is an editable draft on the workflow canvas. Triggers are configured in ToolJet, and running a workflow through MCP is a separate, explicitly authorized action.
What can a workflow step do?
A step can query or write to any connected data source, run JavaScript, branch on a condition, loop over a list, call an AI agent, wait for an approval or return a response to whatever started the run. Workflows use the same 100+ data sources as ToolJet apps.
How do I monitor workflow runs?
Every run is logged step by step, with the status and output of each step. You can see which runs completed, which failed and which are waiting for an approval. Workflows run on the server with configurable timeout and memory limits.
Do workflows run on ToolJet Cloud or self-hosted?
Workflows and AI agents run on self-hosted ToolJet, inside your own infrastructure. ToolJet apps can be built on both ToolJet Cloud and self-hosted ToolJet.

Automate the process your team
still runs by hand.

Start with one real process, and add an approval step wherever you need one.