PreMan simulates agents using your software, finds where workflows fail, and turns those failures into repeatable tests.
npm exec -y premanmcp@latest -- onboardLaunch a swarm, inspect sandbox runs, and turn failures into test cases.
Parallel model testing · Sandbox execution · Repeatable test cases.
See how agents use your software, find where workflows fail, and keep the checks that matter.
preman / simulate · probe · repeatExampleGoalResolve a customer return
ToolsOrders · Returns · Refunds
RuleOpen a return before refunding
CheckCorrect outcome and system state
Bring your tools, workflows, and business rules into a sandbox. Describe what should happen in plain English, including the paths an agent should never take.
Simulate agent traffic to check connection behavior and tool discovery. Then exercise workflow scenarios to see how agents choose tools and act on the results.
Keep the scenario, expected outcome, and evidence behind each failure. Rerun those checks after a tool, model, or product change, and keep expanding your coverage.
Agents choose tools, chain actions, and take paths your team may never have written a test for. PreMan helps you test those paths.
Simulate concurrent agent traffic. Probe connections and tool discovery, with evidence behind every request.
A successful refund can still skip a required return. Test the whole workflow against your business rules.
Discover existing tests, add adjacent scenarios, and rerun the failures you’ve found after every change.
Cursor
Claude Code
CodexWe’re building for a world where millions of agents use software alongside people. See how PreMan can test your product today.
Book a demo