Skip to content

FAQ

Answers about connection, scores, data, and team access.

The CLI is installed. Why is the web app empty?

Section titled “The CLI is installed. Why is the web app empty?”

Installation, sign-in, collection, and evaluation are separate checks. Run prism status, make sure it shows Login: logged in, and, if you record every session with prism setup, that the hooks of Claude Code or Codex are registered, then open a new session and submit a prompt. Look for the new activity in Realtime. A good prism status does not prove that this prompt has arrived or been evaluated. Follow Troubleshooting.

Prism collects activity from Claude Code and Codex. The Claude Code plugin and the Codex plugin work on macOS, Linux, and Windows. The CLI, which is recommended where it is available, can be installed on macOS only. See Install and update for details.

For one session, start Claude Code or Codex with PRISM_OFF=1. To stop for good, run prism setup remove (removes the hooks) or prism uninstall (removes everything the CLI put on your Mac). Stopping does not delete data Prism already received. See Data collection and visibility.

No. The CLI has no key to copy. You sign in with prism login, which shows a code that you approve in the web app. See Quickstart.

Why are there fewer evaluated items than prompts or sessions?

Section titled “Why are there fewer evaluated items than prompts or sessions?”

A subtask groups work toward a goal. It is not one prompt, one model request, or one commit. Some work is pending, excluded, or grouped with related work. Counts also depend on the selected period and table limit. See Scores and work status and Sessions.

No. Counted means the item contributes to the relevant aggregate. Its completion status and Prompt Score answer different questions. Open the item to read the status, evaluation, and inclusion reason.

Does a high score prove productivity or professional skill?

Section titled “Does a high score prove productivity or professional skill?”

No. Prompt Score describes evaluated prompt writing, and the Skill page summarizes observed interaction habits. Neither establishes overall ability, hours worked, or the business value of the result. Read the underlying work and sample coverage before drawing conclusions. See Skill.

No. The CLI prints no scores. Scores appear in the web app, for example in Prompt. prism status shows only sign-in, registration, and sending state.

Can my team manager see my prompts and scores?

Section titled “Can my team manager see my prompts and scores?”

Team screens show values for each member, such as Prompt Score, task states, cost, and usage. No letter grade is shown next to a person, and Team Performance does not read prompt text. The Team report shows neither individual prompt text nor repository paths. Prism also records who opened Team screens, and owners and admins can read that record. Read Data collection and visibility for the roles and screens.

Organization owners and admins can open the Team screens of every Team in the organization. Other members need view access to a Team, which an owner or admin grants in the Team’s settings. See Team access and Teams.

Treat it as an estimate for the recorded usage and available prices. Missing pricing or usage can leave the total partial. A lower cost also does not establish that the same work was completed. See Telemetry and Periods and data states.

Why does an old report differ from the current dashboard?

Section titled “Why does an old report differ from the current dashboard?”

A report retains its generation time, language, period, and included data. The dashboard can later receive more data or show a different range. Compare those conditions first, then generate a new report if needed. See Report.