Know what a change actually takes off your Anthropic invoice.
CostAssured connects read-only to your organisation's usage and cost reports and shows what is driving token spend, what can safely change, and what each change returns at the rates your account is actually billed. Anthropic issues no read-only admin credential, so CostAssured's guarantee here is what our code does and what your audit log shows: reads only, never a write.
Where Anthropic spend drifts
None of this is carelessness. It is what accumulates when usage grows faster than governance, and it is what CostAssured reads for, without writing anything.
Cache writes that are never read back
A 5 minute cache write costs 1.25x the input rate and a 1 hour write costs 2x, so the premium only pays off if something reads the cache back afterwards. A cache nobody reads is pure premium, and CostAssured shows which writes paid off and which did not.
A model with a same-tier successor at a lower list price
Calls keep going to a model after Anthropic ships a same-tier successor at a lower rate. Same tier, lower rate, is a migration that deserves an eval rather than a leap, and CostAssured prices the same volume against the newer rate.
Token usage by key and workspace nobody attributes
Keys outlive the projects that made them, and the invoice only ever shows the total. CostAssured reads token usage by key and by workspace, so a finding names the key and the team behind it.
A cache lifetime chosen by default rather than by traffic
A 1 hour cache costs twice a plain input token to write, where a 5 minute cache costs a quarter more, so the longer lifetime only pays back once two calls have read it. CostAssured prices the 5 minute and the 1 hour writes separately, so a lifetime nobody chose on purpose shows up as the premium it is.
Maximise value, eliminate accidental waste
You chose this platform on purpose and you pay for it on purpose. CostAssured finds the gap between what you meant to buy and what you are actually paying for, then prices every change both ways.
Read-only by design
CostAssured connects with read-only credentials and never writes to your account, on any plan. Every change stays in your hands.
Two numbers on every finding
Every saving is shown as a gross figure and an invoice-impacting figure, each with its basis: your confirmed contract rate, or the rate this account is actually billed.
Safety before savings
Findings are classified against a dependency graph. A resource referenced by an alerting, SLO or detection asset is never treated as low risk, and CostAssured will not recommend removing it.
Remediation you apply yourself
Every finding carries generated remediation artefacts: Terraform, a runbook or configuration JSON, each with a rollback. Generated, never executed.
We store usage metadata and configuration inventory, never telemetry or application content.