Can agents actuallyuse your product?

Watch agents work through your app, inspect the evidence, know what to fix.

Run settings

Who the visitor is when the agent asks for details.

Available in the agent's shell for this run only.

KeyValue

Paste a .env file into any field, or drop one here, to fill several rows at once.

Remote servers the agent can call during this run.

Agent Skills the harness loads for this run: a folder with a SKILL.md, from GitHub, uploaded or dropped here, or pasted.

A headless browser the harness drives through an MCP server. Tracks keeps a screen of each page it shows something new on.

A public GitHub repository checked out in the workspace before the agent starts, at the commit its ref points to now.

Runs use your Cursor plan and appear in your Cursor dashboard.

API key

Custom targetConnect a key first