Overview

Workflows let you chain MCP tool calls into repeatable, parameterised sequences. Define the steps once, run them on demand, track history, and even load-test your server with concurrent executions.

Workflows list — empty state with New Workflow button

The Workflows list. Click New Workflow to create your first workflow, or Import to load one from a JSON file.

Creating a Workflow — Step-by-Step Example

The walkthrough below creates an Echo and Lookup workflow that:

  1. Prompts for a name at runtime using the aaa_echo tool
  2. Passes the echoed value into the who_is tool to look up that person

Step 1 — Name your workflow

Click New Workflow, enter a name, then click Add Step.

New Workflow dialog with name filled in

Give the workflow a clear name. You can also set a default connection to avoid selecting it every time you run.

Step 2 — Configure Step 1 (aaa_echo)

Select the connection (mcp server) and enter aaa_echo as the tool name.

Step 1 configured with mcp server and aaa_echo tool

Step 1 uses the mcp server connection and the aaa_echo tool.

Step 3 — Add a Prompt at Runtime mapping

Click Add Mapping, set the Parameter to echo, and change the Source to Prompt at Runtime. This tells the workflow to ask for the value each time it executes.

Parameter mapping with Prompt at Runtime selected for echo

The echo parameter is set to Prompt at Runtime — the workflow will pop up a dialog asking for this value before executing.

Step 4 — Configure Step 2 (who_is)

Click Add Step, select mcp server and enter who_is as the tool. Then click Add Mapping and set fullname as the parameter with source From Previous Step.

Step 2 configured with who_is tool

Step 2 uses the who_is tool to look up information.

Step 2 fullname parameter mapped From Previous Step

The fullname parameter is mapped From Previous Step — it will automatically receive the output of aaa_echo as its input.

Step 5 — Save and view the workflow

Click Create. The workflow appears in the left sidebar showing its step count.

Saved Echo and Lookup workflow in the detail view

The saved workflow with both steps listed. Click Execute to run it.

Running a Workflow

Click Execute. If any step has a Prompt at Runtime parameter, a dialog appears asking for those values before execution starts.

Runtime Parameters dialog prompting for the echo value

The Runtime Parameters dialog collects any values marked as “Prompt at Runtime” before the workflow begins.

Runtime parameter filled with Garrard Kitchen

Enter the runtime value — here “Garrard Kitchen” — then click Run Workflow.

The execution panel shows real-time step progress and results:

Workflow execution results showing both steps completed

Both steps completed ✓ in 107ms. Step 1 echoed “Garrard Kitchen” and Step 2 received that output as the fullname input — showing the full data pipeline in action.

Parameter Mapping Sources

Each step mapping can draw its value from one of three sources:

SourceDescription
Manual ValueA fixed static value set at design time
From Previous StepThe output of an earlier step (auto-selects the immediately preceding step)
Prompt at RuntimeUser is prompted for this value each time the workflow is executed

Running a Workflow

Click Execute on the workflow detail view. For workflows with Prompt at Runtime parameters, a dialog collects the values before execution begins. The panel then shows real-time step progress, input/output per step, and total duration.

Execution History

Every workflow run is saved automatically. Click History to view past runs including timestamps, duration, and per-step status.

Workflow history showing a completed run

The history panel lists every execution. Click any run to expand the full step-by-step input/output detail.

You can delete individual runs from the history with a confirmation dialog.

Import & Export

Workflows can be exported to a JSON file and imported — useful for sharing workflow templates with a team or backing up your work.

  • Export: Click the Export button on the workflows list
  • Import: Click Import and select a .json file

Load Testing

For each workflow you can run a load test — execute the workflow N times concurrently to stress-test your MCP server:

  1. Open a workflow
  2. Click Load Test
  3. Set the number of concurrent runs and total iterations
  4. Click Start — results show success rate, average duration, and errors

Workflow Tips

  • Use workflows to automate repetitive MCP exploration tasks
  • Combine a search tool with a summarisation prompt for quick research pipelines
  • Use load testing to find throughput limits before deploying to production