Four steps to
objective proof.
Install → Challenge → Score → Share. Each step is designed to capture what matters and ignore what doesn't.
Install. Observe. Ship.
A lightweight MCP server runs alongside your editor. It watches how you work — prompting, iterating, debugging — without ever touching your source code.
The server captures behavioral signals: tool calls, edit patterns, session flow. Think of it as a fitness tracker for how you code with AI.
$ npx goship-mcp start
● MCP server running on localhost:3100
● Observing session… ready.
Real work. Real constraints.
Pick a challenge drawn from the kind of work companies actually need done — building features, debugging production code, extending existing systems.
BUILD
Ship a feature from scratch against a spec and a deadline.
FIX
Diagnose a broken codebase and land a working patch.
EXTEND
Add capabilities without breaking what already works.
Four dimensions. One score.
Your session is scored across four dimensions. Each one measures something different about how you work with AI — not just the output, but the process.
Composite Score
Your code stays yours.
We measure how you work, never what you build. No source code leaves your machine. Period.
Stop talking about what you can build.
Prove it in 15 minutes.