Token Ledger pulls real usage from Anthropic and OpenAI into one cached dashboard, forecasts what's next, flags the days that break the pattern, and lets an agent explain any of it back to you in plain language.
Anthropic and OpenAI usage-reporting keys, encrypted at rest and decrypted only server-side.
A Holt-Winters model projects the next stretch of spend with a confidence band, layered right onto your history.
Any day that jumps well past your rolling average gets flagged on its own, and you can ask the agent to dig into why.
Take a real workload or type in token counts, point it at a different model, and see the monthly difference right away.
A Claude agent with real tool access answers questions about your usage data directly. It queries the data; it doesn't guess.
Ask for a PDF or Excel file on any metric, grouped by day, model, or provider. Built live for that request, not a fixed template.
The agent will get a code-execution tool so it can compute and chart anything from your usage data -- not just today's fixed set of metrics and groupings.
Google Gemini, Amazon Bedrock, and Azure OpenAI, alongside the Anthropic and OpenAI keys you can connect today.
Set a monthly cap and get notified the moment you're on pace to blow past it, instead of finding out at the bill.
A daily or weekly spend summary delivered by email, generated by the same agent that answers your questions today.
Tag usage by project, feature, or customer so spend can be sliced beyond just model and provider.
Shared seats on one account with per-member visibility, instead of today's single-user login.
This is a demo, not a security-audited product. Keys are encrypted via Supabase Vault and only decrypted server-side, and every table is scoped to its owner by row-level security. Still, use a scoped, low-spend key — none of this has been independently reviewed.