Skip to main content
A simulation runs a set of scenarios against a connected agent and records the full conversation transcript for each run. Simulations are the execution layer — they produce the raw data that evaluations then score.

Simulations list

Navigate to Simulations in the sidebar to see all simulations in your organization. The table shows:
ColumnDescription
Simulation NameThe name you gave the simulation
AgentThe agent it ran against
Scenario GroupsThe groups the selected scenarios belong to
Created ByThe team member who created it
Last RunWhen the simulation was created
StatusCurrent status (see below)

Status colors

StatusColor
Completed Green
Running Blue
Failed Red
Use the Filter button to narrow results by agent, creator, or status. Use the Search bar to filter by simulation name or agent name.

Creating a simulation

Click New Simulation to open the creation sheet.

Prerequisites

Before creating a simulation you need:
  • At least one connected agent (see Agents)
  • At least one scenario (see Scenarios)
If no scenarios exist, the sheet will prompt you to go to the Scenarios page first.

Simulation form

Agent — Select the agent to run the simulation against. If you have not connected any agents yet, an inline banner links to the Agents page. Name — A descriptive name for this run (e.g. “Insurance Bot v3 - Home Deductibles - May 20”). Description — Optional notes about what this simulation is testing or what changed. Scenarios — Select the scenarios to include. Scenarios are grouped by their scenario group in the selector. Use the search input and “Created by me” filter to narrow a long list. Click a scenario name to preview its goal, profile, and knowledge base before selecting. Conversations per scenario — How many independent conversations Arkdock should generate for each selected scenario. Default is 1. The maximum per scenario is dynamic: the total conversation count across all selected scenarios is capped at 50, so the per-scenario limit decreases as you select more scenarios. Higher counts produce more variation and statistical reliability but take longer to run. Max turns per conversation — The maximum number of exchange turns before a conversation ends. Default is 5. Maximum is 10. Set this to match the typical interaction length for your agent.

Running the simulation

The Run Simulation button is disabled until an agent is selected, a name is entered, and at least one scenario is selected. Click the button to start. You are navigated directly to the new simulation’s detail page.

Simulation detail page

  • Simulation name and status badge
  • Agent name, creation date, and conversation count

Conversations table

Each row is one conversation generated during the simulation.
ColumnDescription
ScenarioThe scenario ID that generated this conversation
SummaryA short label derived from the scenario goal
GoalThe full goal text from the scenario
TurnsNumber of exchange turns in the conversation
StatusCompleted, Running, or Failed
Click any row to open the Conversation Modal and read the full transcript.

Conversation modal

Use the Previous and Next arrows to navigate between all conversations without closing the modal.

Cancelling a simulation

Simulations that are queued but have not started yet can be cancelled. Click the kebab menu on a simulation row with Pending status and select Cancel. A toast confirms the cancellation. Simulations that are already running cannot be cancelled mid-execution.

Rerunning a simulation

Click the kebab menu on any simulation row and select Rerun. The rerun sheet pre-fills the simulation name and shows a summary of the original configuration (agent, scenario count). Edit the name if you want to distinguish this run from the original, then click Run Simulation. The rerun creates a new simulation record with a new ID. It uses the same agent and the same set of scenarios but generates fresh conversations.

Deleting a simulation

Click the kebab menu and select Delete and confirm. The simulation and its conversations are deleted. Evaluations that reference this simulation will note that the source simulation was removed.

FAQ

Run time depends on the number of scenarios, conversations per scenario, max turns, and your agent’s response latency. A simulation with 10 scenarios, 1 conversation each, and 5 turns typically completes in a few minutes.
A conversation is marked Failed when Arkdock could not complete it — for example, if the agent endpoint returned an error, timed out, or produced an unparseable response. Failed conversations are excluded from evaluation scoring but remain visible in the transcript view.
Yes. Simulations run independently and do not interfere with each other.
Completion rate is the percentage of conversations that reached a Completed status. Conversations with a Failed status are counted as incomplete. Running conversations are excluded from the calculation until they finish.
Create a new simulation after updating your agent. Rerun is best used to get more conversation samples from the same agent configuration. Running a fresh simulation preserves a clear record of which agent version was tested.