> ## Documentation Index > Fetch the complete documentation index at: https://docs.uservox.ai/llms.txt > Use this file to discover all available pages before exploring further. # Lead Simulator > Step a sandbox lead through a process node by node, choosing call outcomes at each step, without making a real call or touching real leads. This is a different feature from [Simulate a Voice Agent](/essentials/simulation), which tests an agent's conversational responses against sample variables in the "Try" tab. The Lead Simulator instead walks a **process** — nodes, conditions, dispositions, and scheduling — one step at a time. ## Overview The Lead Simulator lets you dry-run a process against a throw-away sandbox lead: pick an outcome for each call step (Connected / Not Connected → a disposition), and watch which node the process moves to next. It's the fastest way to check a process's branching logic — conditions, disposition routing, call windows — without waiting for real calls. *** ## Starting a Session Open **Simulator** from the app (or a process's **Simulate** action, which pre-fills the Process ID). Fill in: * **Process ID** (required) — the process to simulate. * **Project ID** (optional) — defaults to the process's own project. * **Start state** (optional) — defaults to the process's actual initial state. * **Lead name** / **Phone** — defaults to a clearly-fake sandbox lead (`Simulator Lead`, `9999999999`). * **Metadata** / **learnedData** (JSON, optional) — seed values the process can read, useful for testing condition branches that depend on lead fields. * **Clock** (optional) — sets a fixed simulated time for deterministic replay; leave blank to run in real time. Simulator new session form, with process ID, sandbox lead defaults, and how-it-works notes The sidebar's **How it works** notes explain the mechanics: the session uses a throw-away sandbox lead (`__sandbox: true`) deleted when you close the session; every `call` action pauses for you to pick an outcome, while non-call actions (like `transform` or `webhook_call`) run in dry-run mode automatically unless a downstream step needs a real value. Sessions are persisted as `ProcessTestRun` records with `mode='SIMULATOR'`. Click **Start session** to open the workbench. *** ## The Workbench Simulator workbench: visited states, current state with outcome picker, lead snapshot, and ledger * **Visited states** (left) — the path taken through the process so far. * **Current state** (center) — what node the simulation is on, and what it's about to run. For a call step, choose **Connected** or **Not Connected**, then pick a specific disposition, grouped by outcome type (Success / Neutral / Failure): Outcome picker showing process-specific dispositions grouped by Success, Neutral, and Failure Use **Skip tick** to advance without running the step, or **Fast-forward** to run several steps automatically. * **Lead snapshot** (right) — the sandbox lead's current phone, state, and last disposition, with raw metadata/learnedData available to expand. * **Ledger** (right) — a chronological log of every job and action the session has run. After picking an outcome, review the raw payload for the next step and click **Advance** to run it. *** ## Ending a Session Click **Close session**. You'll be asked to confirm — closing deletes the sandbox lead, so nothing persists beyond the `ProcessTestRun` record of what happened. ## See Also * [AI Process Assistant](/guides/process/ai-process-assistant) — propose process changes via chat * [Simulate a Voice Agent](/essentials/simulation) — test an agent's conversation, not a process's routing