Private AI spend intelligence

Usage Radar turns Claude Code, Codex, Gemini and provider billing into one private cost workspace—without uploading your prompts or source code.

Connect System Logs Explore Live Workspace

Estimate My Spend

Local-first
No raw prompts required
No ads or data selling
Free to start

usage-radar / overview SAMPLE WORKSPACE

MONTHLY AI SPEND · USD$438.12

14.2% optimized

Cache saved · USD$61.40

Sessions184

Active models7

30-day spend

Jun 21Jul 5Jul 20

Opus$254.11

Sonnet$140.20

Haiku$43.81

Built for the tools already in your terminal

Claude CodeCodexGemini CLIGitHub Copilot

One source of cost truth

Stop guessing. Start governing.

Usage Radar gives every number a source, every trend a reason, and every optimization a dollar value.

Unified spend

One ledger across agents and models.

Compare API-equivalent cost, provider billing, and local agent usage without merging unlike data.

Opus58%$254.11

Sonnet32%$140.20

Haiku10%$43.81

Cache economics

$61.40

saved by cache reads this month

Turn invisible token reuse into a concrete efficiency signal.

60.6% prompt cache hit rate

Privacy boundary

Your code and prompts are not the product.

The local connector exposes aggregate usage metrics over loopback. Raw prompt content is never required for the dashboard.

Local log processing
Revocable provider keys
Usage telemetry only

Five deterministic stages

From “probably fine” to fully understood.

Detect

Link only the Claude Code, Codex, or Gemini CLI folder you choose.

Privacy

Extract usage counters locally. Prompt and response text never enters the normalized record.

Normalize

Map every source to one deterministic token, cache, model, session, and project schema.

Intelligence

Surface anomalies, cache efficiency, forecasts, and the sessions driving spend.

Reconcile

Compare observed API-equivalent cost with provider totals without pretending estimates are invoices.

See exactly how the data flows

No setup required

Put a number on your usage in under a minute.

Start with transparent API-rate arithmetic. When you want exact local usage, connect supported system logs in your browser.

No account needed. This estimate runs entirely in your browser.

Hours of agent use a day4h

Model you lean on

Your estimate:$202.40/ month · API-equivalent

The same work on Sonnet 5 would run about $81 — a difference of $121 a month.

Rough: assumes a heavy agentic hour (800k cache-read, 80k input, 60k output tokens) across 22 working days, at published API rates. It is not your data. Connect the local connector for the real figure.

Designed around a hard boundary

Financial clarity without surrendering your work.

Aggregate counters power the product. Your source code, prompt text, and raw conversations do not need to leave your machine.

Read the privacy model

Built for one developer or a whole cost center

Private for individuals. Accountable for teams.

INDIVIDUAL

Understand the economics of your own agent workflow.

See model mix, cache savings, anomalies, and provider deltas without handing over your work.

Connect system logs

TEAM

Give engineering and finance one defensible ledger.

Shared budgets, role-aware rollups, attribution, alerts, exports, and an organization audit trail.

Compare Team

Clear commercial boundaries

Start free. Upgrade for continuity, not access to your own data.

The complete local dashboard is available now. Paid capabilities are documented for private launch; checkout remains closed.

Free$0Local dashboard

Local Plus$9lifetime · private launch

Pro$9/mo$79 annually · private launch

Team$19per active seat · 3 minimum · private launch

See the full comparison

Security and privacy commitments

Hard boundaries, written down.

No prompts or source code by default. Local processing is the baseline, not a premium feature.
Sync requires explicit consent. Privacy-max and cloud-intelligence modes disclose different tradeoffs.
Billing never grants access from a redirect. Signed, idempotent Stripe webhooks control entitlements.
Your data is exportable and deletable. Legal retention exceptions are described before deletion.

Read the security model

Frequently asked

The important questions, answered plainly.

Does Usage Radar upload my prompts or source code?

No. The local connector reads usage fields and exposes aggregate metrics over loopback. Raw prompt text and source code are excluded by default.

Is the free product actually free?

Yes. One-machine local use, 30-day history, estimates, local ingestion, and CSV export require no card.

What is the difference between Local Plus and Pro?

Local Plus is lifetime local access. Pro adds opt-in encrypted sync, unlimited history, alerts, reports, reconciliation, and up to three devices.

Can the cloud read encrypted usage?

Not in privacy-max mode. Cloud-intelligence mode separately exposes only the aggregate fields you approve, and that consent can be revoked.

Your usage already has a story

See where every dollar went.

Start locally without a card, then upgrade only when the paid capabilities are useful.

Connect System Logs Compare plans

Know exactly what yourAI coding agents cost.