Model calls
Which model answered, under which policy and whether a fallback was used.
Platform
Audit Ledger
Every governed step, tool call and approval lands on a hash-chained ledger for your tenant. If anyone changes a past entry, verification says exactly where. Try it below.
How the chain works
Tamper evident, in three ideas.
You do not need to trust our word that a record is unchanged. The maths shows it.
Each entry is fingerprinted
When Spine writes an entry it computes a hash: a short fingerprint of the contents. Change one character and the fingerprint changes completely.
Each fingerprint includes the one before
The hash of every entry is computed over the previous entry's hash plus its own contents, so the entries form a chain.
Verification walks the chain
Recompute every fingerprint from the first entry. If any past entry was edited, the chain stops matching at that exact entry.
Interactive sandbox
Tamper with a ledger and watch it get caught.
Twelve sample events from a fictional collections run. Verify the chain, change an amount or a decision in one row, then verify again.
Sample data, in your browser only
This demo is keyless and runs entirely on this page with your browser's own SHA-256. The production ledger uses a keyed hash and a separate chain per tenant. Nothing here reads or reports on any live ledger.
Sample dataA fictional tenant. Runs only in your browser.
Sample ledger ready. Press Verify chain to check every link.
Building the sample chain...
What gets recorded
The whole decision, not just the outcome.
When a regulator or a customer asks why something happened, the answer is already written down in order.
Guardrail results
Every check that passed or blocked, with the rule that fired.
Tool calls
What the agent asked a system to do and what came back.
Approval decisions
Who decided, which of the four choices they made and their reason.
Questions
The audit ledger, answered.
No. It runs on a fictional sample ledger in your browser and sends nothing anywhere. It uses the same idea as the product so you can see how tampering is caught.
Production entries are chained with a keyed hash, so forging a valid fingerprint needs a secret key as well as database access. Each tenant has its own chain, isolated in the database.
The application role that writes the ledger cannot update or delete entries. Any change made around it breaks verification at the changed entry.
Your team can run it from the console at any time and auditors can be given the evidence pack. The result names the first entry that fails, if one does.
See the ledger behind a real governed call.
We will run a call, open its entries and verify the chain in front of you.