Why the provider dashboards aren't enough
OpenAI and Anthropic show you a total and a usage graph. What they do not show is the split that decides the bill: how much is input (billed at full price, often uncached) versus output, which model is driving growth, and which calls went to a frontier model a cheaper one would have handled. Token prices keep falling, but agents use 5 to 30 times more tokens, so the total climbs anyway.
What nable shows
Cost by model
Spend per model across OpenAI, Anthropic, and Bedrock, so you see which model is the growth driver, not just the account total.
Input, output, and cache
The split that matters. Most of a bill is input tokens at full price; nable shows how much prompt caching would recover.
Cheaper-model candidates
Calls sent to a frontier model that a cheaper one would answer the same, with the monthly saving.
# after `uvx nable`, with your model-provider key
You: What is my AI bill by model, and where can I cut it?
nable: $9,400/mo. Opus 62% · GPT-4o 24% · Haiku 14%.
89% is input tokens, uncached. Caching the shared context
saves ~$1,900/mo. 31% of Opus calls fit Haiku, ~$1,400/mo.See your real AI bill, split by model, next to your cloud spend. nable reads your OpenAI, Anthropic, and Bedrock usage locally, flags the calls a cheaper model would answer the same, and shows what caching would save. Free for solo use, your usage never leaves your machine.
Common questions
How do I track OpenAI and Anthropic API costs together?
nable reads your OpenAI, Anthropic, and Bedrock usage and shows the token bill split by model, by input vs output, and by workload, in one place, next to your cloud spend. Ask in Claude or Cursor after connecting a model-provider key. For org-level totals, an admin/org key is needed; nable tells you if a regular key returns no billing data.
Can nable show token spend by model?
Yes. nable breaks AI spend down per model rather than just the account total, so you can see which model is driving growth and route the calls a cheaper model would answer the same.
Does prompt caching actually reduce cost?
Yes. Cached input tokens are billed at a steep discount (often ~90% off). Most of a typical bill is input tokens at full price, so nable shows how much of your input is cacheable and the monthly saving.
Does my API usage leave my machine?
No. nable runs locally and reads your provider usage read-only. Your keys stay in your OS keychain and your usage data is not sent to a nable backend.