AI Governance

Prompt IntelligenceNew

Captures and analyses every Claude Code session across your development organisation — surfacing prompt quality issues, token costs, model usage, and per-developer activity without requiring any per-developer setup.

Get a Demo Login to App

Prompt Intelligence

Visibility and governance over your AI developer tools

What it does

Key capabilities

Automatic Session Logging

Claude Code Stop hook logs every session automatically on close — no agent action or developer setup required.

Per-Developer Visibility

Track prompt volume, token usage, and cost per developer across the organisation.

Token Cost Analytics

Real-time cost tracking per model, per project, and per user — with trend analysis over time.

Bad Prompt Detection

Automatically flags prompts that are too brief, overly long, or missing clear action verbs — with coaching feedback.

Tenant Isolation

Each client's data is scoped via their unique Tenant ID — Strategix data never mixes with GIC data.

Managed Settings Integration

Deploy org-wide via the Anthropic Managed Settings portal — set once, applies to all 150+ developers automatically.

Highlights

Why Prompt Intelligence?

  • Zero per-developer setup — configured once by admin
  • Deployed via Anthropic Managed Settings portal
  • Automatic session logging on every Claude Code close
  • Cost breakdown by model, project, and developer
  • Bad prompt detection with remediation coaching
Request Demo Talk to us
Included in OneView

Prompt Intelligence is part of the Hakware OneView platform. All findings, events, and data from this module flow into your central dashboard alongside every other security signal in your environment.

24/7 Support Available
Phone: +27 060 984 1210
FAQs

Frequently asked questions

Prompt Intelligence uses a Claude Code Stop hook that fires automatically when any Claude Code session closes. The hook logs session metadata — token counts, cost, conversation summary, prompt quality flags — to the Hakware platform via the organisation's tenant ID. Deployment is through the Anthropic Managed Settings portal, requiring a single admin configuration for the entire organisation.

The platform evaluates prompts against three quality signals: brevity (prompts under approximately 20 words often lack sufficient context for accurate results), excessive length (prompts over 2,000 words may indicate knowledge-dumping rather than structured requests), and missing action verbs (prompts without a clear instruction verb are harder for the model to act on effectively).

Every session log includes the developer identity (from the Claude Code environment), the project directory context, the model used, and input/output token counts per model tier. Prompt Intelligence aggregates these into cost dashboards showing spend per developer, per project, per model, and over time — enabling AI cost allocation and budget forecasting.

No. Prompt Intelligence collects metadata — token counts, cost, session duration, quality flags, and summary signals — but does not store the actual prompt and response content. This privacy-first approach means you get visibility into AI usage patterns and cost without capturing developer intellectual property or sensitive project context in the platform.

Related modules

AI WatchNew

Workstation AI tool usage monitoring and governance

Learn more

AI AuditorNew

AI governance, compliance, and risk auditing

Learn more

See Prompt Intelligence in action

Request a personalised demo and we'll show you exactly how Prompt Intelligence works within your environment.

Request Free Demo All Modules