Skip to main content

Scorecard from the command line

AI evaluation testsets, metrics, runs and scores. GBrain gives Scorecard, and every other application you connect, a command line.

Get Personal, $49 a month

One price for one person: Scorecard and every other application GBrain connects, from a terminal or any AI. Cancel any time.

Install the GBrain command line, then connect Scorecard

Scorecard is connected on its own sign-in page, in the browser. A key or a sign-in never passes through the terminal.

# AI agents: read gbrain.io/llms.txt
curl -sSL https://gbrainio.terminalwire.sh | bash
gbrainio terminal connect

The installer is a shell script for macOS and Linux. Read it before you run it.

# Every Scorecard command, once Scorecard is connected
gbrainio tool help scorecard

What GBrain can do with Scorecard

Scorecard is an evaluation platform for AI products: a team keeps its testsets, metrics and runs in one place and scores model output against them. Connected to a workspace, GBrain can answer “how did last night’s run score on the support testset?”, turn examples from a thread into test cases, or set up a new metric.

  • Create and update testsets and test cases
  • Define metrics and record runs and scores
  • Review runs, records and annotations
  • Search Scorecard's documentation

What a terminal may do with Scorecard, and what GBrain keeps

Off or On, for each terminal

Each terminal reaches Scorecard at the level set for it in the browser: Off, or On, which is every command Scorecard's server lists. A terminal reads its level and never raises it.

1,000 actions a day for each command

Each of Scorecard's commands stops at 1,000 actions a day on the account it reaches, whichever terminal or AI asks, and starts again at midnight UTC. Only that command stops.

Every action on one record

Each command a terminal runs on Scorecard is a row on GBrain's activity log: which terminal, which account, what it did and whether it was allowed, refusals included. gbrainio activity ls --tool scorecard reads it back.

From a terminal, Claude Code or Codex

Scorecard's own server lists its commands only to a terminal that has connected it, so they are not on this page. Once Scorecard is connected, gbrainio tool help scorecard says what each of them does, and gbrainio tool ls lists them with everything else the terminal can run. Claude Code and Codex run the same commands, at the same level, on the same record.

From Scorecard

GBrain reaches Scorecard at its own server, mcp.scorecard.io. On Scorecard's own pages: its website, how the connection works, its terms of service and its privacy policy.

The GBrain Tools command line works with every other application GBrain connects, each with a page like this one.

Every application's command line

Scorecard from a terminal, with GBrain Personal

Personal is $49 a month for one person. Connect Scorecard in the browser, and say how far each terminal and each AI may go with it. Change any of it, or take all of it back, whenever you want.