Your team's LLM usage is scattered across API dashboards, coding tools, and cloud consoles — each with its own model names and units. The AI Explorer normalises all of it into one lens: usage by model, project, and user, measured in tokens and API-equivalent value.
On the Business plan — free to try during your 14-day trial.
Usage by model · last 30 days
$8.2k ≈ API value · 1.9B tokens
claude-sonnet-5
Claude Code · 812M tokens
$3.4k
gpt-5
OpenAI · 414M tokens
$2.9k
gemini-2.5-pro
Vertex · 512M tokens
$1.4k
gpt-5-mini
OpenAI · 162M tokens
$0.5k
claude-sonnet-5 · Input 61% · Output 9% · Cache read 27% · Cache write 3% — blended ≈ $4.15 / 1M tokens
Illustrative example.
Direct answer
The AI Explorer is a cross-provider view of LLM usage: every model your team uses, from every connected provider and tool, normalised into one lens and measured in tokens and API-equivalent dollar value. Group it by provider, model, project, or user; drill into input, output, and cache token splits per model; and export any view. Estimated and billed usage stay honestly separate throughout.
Raw model strings and billing SKUs resolve to one standardised base model, so the same model used via an API, a coding tool, and a cloud platform rolls up to a single row — across OpenAI, Anthropic and Claude Code, Cursor, Google Vertex and Gemini, Grok, and more.
Group usage by provider, model, project, or user; filter any combination; chart it as stacked bars, lines, or area; and pivot it into a daily, weekly, or monthly table you can export as CSV or JSON.
Expand a model to see its input, output, cache-read, and cache-write token split, billed dollars per direction where the vendor bills that way, and a blended effective rate per million tokens at your actual mix.
Estimated and billed usage are kept separate and deduplicated, unpriced usage never shows a fake $0, per-provider coverage flags mark exactly what data exists, and non-generative models stay out of your token totals.
Dollar figures in the AI Explorer are API-equivalent value: what the usage would cost at API list rates. That makes subscription-tool usage comparable with pay-as-you-go API spend in one number — but it is not an invoice, so billed cost is always shown separately, estimated and billed rows are deduplicated, and anything without a known price shows tokens with no dollar value rather than a fake zero.
When you're ready to act on what you see, Model Recommendations take the same usage mix and find cheaper equal-or-better alternatives.
The raw unit — comparable across every provider, immune to price changes.
Usage × list rates. One number that makes Claude Code, Cursor, and API usage comparable.
What the vendor actually charged. Always shown separately, never blended.
Connect your AI providers and tools, and the AI Explorer shows usage by model, project, and user — in one lens, with honest numbers.