# Testing (/dashboard/testing)

The **Testing** tab runs automated test suites against your server. A suite describes scenarios, prompts and expected behavior, and Manufact executes them across different MCP clients and LLM models, because a server that works in one client can still fail in another.Test suites require the [Hobby plan or higher](/dashboard/billing). Eval runs are billed at [$1 per run](/dashboard/billing).![Testing suites list](/screenshots/dashboard/testing/testing-suites.png)## Creating a suite [#creating-a-suite]1) Go to **Testing** and click **Add test suite**.
2) Define the scenarios to test: what to ask, and what a correct response looks like.
3) Pick the clients and models to run against, **Inspector** and **ChatGPT** are available today, with Claude and Cursor coming soon.## Running and reading results [#running-and-reading-results]Run a suite on demand from the list. Open a suite to see run history with pass/fail per client and model, regressions that only affect a specific client or model stand out immediately.![Suite run history with per-client results](/screenshots/dashboard/testing/suite-run-results.png)![LLM judge result drawer](/screenshots/dashboard/testing/suite-results-drawer.png)## When to run tests [#when-to-run-tests]- **Before publishing**: alongside the [Checklist](/dashboard/publish-checks), to make sure behavior matches your tool descriptions.
- **After changing tool schemas**: renamed tools and changed parameters are the most common source of silent client breakage.
- **When upgrading frameworks**: verify your MCP framework upgrade didn't change wire behavior.Testing works for both Manufact-hosted and external servers.