Skip to content
Hivemind Receipts Verifiable Agent Evidence
Open console
Signed, logged and witnessed records

Signed, logged and witnessed records of agent actions.

When an AI agent calls an API, moves funds, or updates a record, Hivemind creates a signed receipt (.hvr). Every tool call can be evaluated against your policy, recorded in an append-only transparency log, and verified locally in your browser or with the hvr CLI.

Verify a receipt Open console
Apache-2.0 browser and hvr CLI verifier · Witness nodes w0..w2 operated by Hypereum · External witnesses: none yet · Read the technical specification
Sample · Refund receipt Verified · T1
Agent
billing-agent acting for jane@acme.com
Operation
POST api.stripe.com/v1/refunds ($240.00)
Policy
refund-policy · v3 · approver: finance-lead
Witnesses
3 cosignatures, 2 required (w0..w2 operated by Hypereum · External witnesses: none yet)
Outcome W-level
W2: Counterparty-confirmed · Signed webhook, DKIM/ARC email receipt, or XMLDSig-signed bank confirmation (camt.054 / pain.002).
Run ID: computed by WASM
Why teams use Receipts

Why teams use Receipts

Application logs can be edited or lost when an incident happens. Receipts give your security, finance, and operations teams portable records that anyone can check offline.

01 · Tamper-evident history

Know what your agents actually did

Every tool call is signed and appended to an ordered log. Sequence numbers make missing or deleted actions immediately visible.

02 · Policy & approvals

Evaluate before execution

Every tool call can be evaluated against your policy before it runs. High-impact actions pause for human approval, and both allowed and blocked actions emit a receipt.

03 · Privacy by default

Keep private data off the public log

Only salted commitments are written to the public log. You can share a receipt and selectively disclose specific fields to an auditor without exposing customer data.

How it works

How it works

01

Wrap your agent tools or gateway.

Add the Receipts wrapper to your MCP server, agent framework, or API gateway in one command without rewriting your prompts.

02

Evaluate rules and record a signed receipt.

Each action is checked against your workspace rules, signed by the agent's workload key, and recorded in the append-only log.

03

Verify anywhere, offline or in your browser.

Drop any .hvr file into the browser verifier below or run hvr verify in your terminal with zero network calls.

Four questions every receipt answers

What a receipt answers

Question 01

Who did it?

Identifies the specific AI agent workload, its attestation tier (such as T1), and the human user or service that delegated the task.

Question 02

What happened?

Records the tool name, target service, timestamp, sequence number, and a commitment to the exact request and response parameters.

Question 03

Who allowed it?

Captures the policy rule that evaluated the call and, when human sign-off was required, the passkey approval attached to the action digest.

Question 04

Did the outside world confirm it?

Reports the external confirmation level from W0 (dispatched) to W3 (W2 evidence plus a qualified electronic seal or qualified timestamp over that evidence). See the full scale on Spec.

Local WebAssembly verifier

Verify a receipt.

Drop any Hivemind receipt file (.hvr) below. Your browser checks the signature, log inclusion proof, and witness cosignatures locally on your device.

Drop a .hvr receipt file here
Runs 100% in your browser using pinned sample log trust roots by default.
Advanced: custom checkpoint and trust roots

Or try one of the sample receipts below:

No receipt loaded yet. Drop a .hvr file above or choose one of the sample receipts below the drop zone.