--- title: Test a template before anyone runs it · Tallyfy Pro description: Test a Tallyfy template before anyone runs it. Ask Claude or any connected AI to walk every path through a template and report the problems it finds, like a step no answer can ever reach or two rules fighting over the same step, without launching a single process or involving a single person. lastUpdated: 2026-09-09T19:42:45.000Z source_url: html: https://tallyfy.com/products/pro/integrations/mcp-server/template-testing/ md: https://tallyfy.com/products/pro/integrations/mcp-server/template-testing/index.md --- ## Find the broken path before a person does Until now the only way to find out what a template really does was to launch it, wait for real people to answer, and watch something break. If a step could never appear, you learned that when somebody’s onboarding stalled. Two tools on the Tallyfy MCP server change that. Ask a connected AI to test a template and it walks every path through it internally, then tells you what is wrong. Nothing is launched. Nobody is assigned. No task is created. Both tools only read They read one template and hand back a list of problems. Neither can change your template, launch a process, or touch anyone’s tasks. ### What to say You do not need to name the tool. Ask in plain language: * “Test my Employee Onboarding template and tell me what is broken.” * “Is there any step in this template that can never be reached?” * “What happens in this template if someone picks Decline on the first form?” The last one is the scenario version. You describe the answers a person might give, and it tells you which steps that person would actually see. ### What it reports, and what it does not It reports **problems, not paths**. A template with a few forms can have thousands of possible routes through it, and a list of those helps nobody. You get the handful of things that are wrong. Twelve kinds of problem, in plain terms: | It tells you | Why that matters | | - | - | | A step no answer can ever reach | The commonest one. Usually a rule hides it and nothing ever shows it again. | | A rule that can never fire | It waits for a combination nobody can produce, so it is dead weight. | | A rule that fires but changes nothing | Showing a step already visible, or hiding one already hidden. | | Two rules fighting over the same step | One shows it, one hides it, and the result depends on the order. | | A step nobody can finish | The process gets stuck there, because no path through it ever completes. | | A step with nobody assigned to it | It is visible, but only assignees may complete it, so it stalls. | | A rule pointing at something deleted | The step or field it names is gone. | | A deadline counting from a step that never happens | The deadline can never be worked out. | | A rule waiting for an answer the field no longer offers | Somebody edited the dropdown and left the rule behind. | | A start date that is set but ignored | The step is on “start anytime”, so the date does nothing. | | A rule whose test is always true | It compares a field against every value the field offers, so it never filters anything. | | A rule nothing ever asks | No step and no other rule ever triggers it, so it sits there unused. | **What it cannot tell you** is whether the process is a good idea. It checks that the template can work, not that it should. It also cannot see anything outside Tallyfy, so a step that calls a webhook is treated as doing something it cannot inspect. ### On very large templates A template with many form fields has too many combinations to try all of them, so on those it tests a large sample instead and **says so in the answer**. When that happens it will tell you what it found rather than claiming a step can never be reached, because it did not check every path. ### Getting set up You need the Tallyfy MCP server connected to your AI tool first. That is a one-time setup, and each tool has its own page: * [Claude](https://tallyfy.com/products/pro/integrations/mcp-server/claude-anthropic/) * [ChatGPT](https://tallyfy.com/products/pro/integrations/mcp-server/openai-chatgpt/) * [Google Gemini](https://tallyfy.com/products/pro/integrations/mcp-server/google-gemini/) * [Microsoft Copilot Studio](https://tallyfy.com/products/pro/integrations/mcp-server/microsoft-copilot-studio/) Once connected, testing a template needs nothing else switched on. ### Where this came from Templates that use several automations together can end up with a step no combination of answers ever shows. Nothing in the editor flags it, and the usual way to find out is that a live process stops halfway and somebody has to work out why. Customers asked for a way to check a template before running it, and this is that. ## Related articles [**Integrations > MCP server**](https://tallyfy.com/products/pro/integrations/mcp-server/) Tallyfy’s MCP Server lets you control workflows through plain English in ChatGPT, Claude… [**Use Cases > Audit and improve your templates**](https://tallyfy.com/products/pro/integrations/byo-ai/use-cases/audit-and-improve-your-templates/) Tallyfy’s AI can audit any process template for missing instructions, form fields, assignees… [**Byo Ai > Use cases**](https://tallyfy.com/products/pro/integrations/byo-ai/use-cases/) Connect any major AI platform to Tallyfy through MCP to manage tasks, check process status… [**Mcp Server > Using Tallyfy MCP server with ChatGPT**](https://tallyfy.com/products/pro/integrations/mcp-server/openai-chatgpt/) ChatGPT Enterprise and Team users can connect to Tallyfy’s MCP server through OAuth 2.1 to… ## Was this helpful?