Skip to content

Search is only available in production builds. Try building and previewing the site to test it out locally.

Core Concepts

Optra Prism is built on a few core ideas.

AI coding agents are powerful but opaque. Developers face:

  • No prompt feedback — you don’t know if your prompts are efficient or wasteful
  • Invisible throttling — rate limits silently slow you down with no visibility
  • No cost visibility — token spend accumulates with no breakdown by session, model, or pattern
  • No coaching — you repeat the same mistakes because nothing tells you what to improve
  • No guardrails — no budget caps and no policy enforcement

Prism solves these by instrumenting the AI coding workflow and surfacing insights.

A good prompt — and a good sub-session — reaches the goal using the minimum necessary tokens, turns, human think time, and response latency, given the current session context.

Those four fundamental values drive every score Prism computes.

Prism produces three personal scores, each answering a different question:

ScoreQuestionUnit
SpeedHow much focused AI-coding time do you put in?Hours/week
SkillHow well do you direct the AI and recover from mistakes?0–100
EfficiencyHow many tokens does each active hour cost?Tokens/hour (lower is better)

The scores are deliberately not collapsed into one number — see PRISM Scores for why and how each is computed.

Prism’s data model is three measurement layers plus a parallel intelligence pipeline:

  • Layer 0 — Telemetry. OpenTelemetry from Claude Code.
  • Layer 1 — Measurement. Metrics · Prompt Efficiency Score · Sub-Session Efficiency Score.
  • Layer 2 — Prism Score. Speed · Skill · Efficiency.
  • Insight. Intent-adaptive agent pipeline that scores each prompt feeding Layer 1.

See Architecture for the full picture.

Prism creates a feedback loop between coding and coaching:

Code with AI → Capture telemetry → Score & analyze → Surface insights → Improve → repeat
  • Real-time: the plugin advisor nudges you on each prompt before submission
  • Session-level: the engine scores each sub-session and detects waste patterns
  • Trend-level: the dashboard shows improvement over days and weeks
  • Recommendations: data-driven suggestions for model rightsizing, prompt patterns, and budgets