EuthyneKnow which agent did what. Prove it.
Every action your agents take becomes a signed, tamper-evident receipt. End-to-end encrypted, verifiable by anyone.
The Euthyne receipts console
EuthyneFileEditViewMerged PR #482 into main
Approved by policy prod-deploy, 3 checks passed.
Issued refund of €240.00
Customer #8812, ticket SUP-1934, under the €500 limit.
Read customer record #8812
Scope: billing history, no payment data.
Called claude-opus-5-5 via OpenRouter
Prompt and output hashed, 4,812 tokens.
Rotated staging API key
Old key revoked, new key sealed.
Flagged: refund over limit
€1,200 refund blocked, escalated to a human.
Merged PR #482 into main
deploy-bot
for Acme Inc · 9:41 AM
Verified by Euthyne
Signed by deploy-bot (did:verana:acme:deploy-bot), acting for Acme Inc. Hash chain intact, policy prod-deploy allowed it. No action needed.
Action: git.merge on acme/api, pull request #482 into main.
Inputs and outputs are hashed into the receipt. The content itself stays encrypted in your infrastructure.
Previous receipt rcpt_7f3a…c21. Next receipt rcpt_9b10…4e7.
— Signed 2026-10-06 09:41:07 UTC
receipt-482.jsonReplay any incident
down to the action.
When an agent misbehaves, Euthyne shows which agent did it, for whom, under which policy, and proves the record was never edited.
Today · 1,284 actions sealed
- deploy-bot — PR #482 merged
- billing-agent — refund €240
- billing-agent — refund over limit
- support-agent — scope exceeded
- legacy-crawler — key revoked
- test-agent — identity expired
- Audit pack Q3 · Incident 0417
Proof, not logs.
Basic
Self-signed
Your SDK signs every receipt. For teams starting to audit their agents.
- Tamper-evident receipts
- Content stays in your infra
- Public verification page
- Agent identity via Verana
- One-line SDK
Witnessed
Countersigned
Your gateway, our signature. For teams whose clients or auditors ask for proof.
- Tamper-evident receipts
- Content stays in your infra
- Public verification page
- Agent identity via Verana
- Independent countersignature
Notarized
Full proof
Our gateway in the path. For regulated teams that need completeness, not just integrity.
- Tamper-evident receipts
- End-to-end encrypted content
- Public verification page
- Independent countersignature
- Completeness guaranteed
01What is Euthyne?
Euthyne is a forensic audit layer for AI agents. Every action an agent takes becomes a signed, tamper-evident receipt that records which agent acted, on whose behalf, with which model and under which policy. Anyone can verify a receipt.
02How do I integrate it?
Point your agents' model calls at the Euthyne gateway by changing one base URL, such as ANTHROPIC_BASE_URL or OPENAI_BASE_URL, or sign receipts in your own code with the SDK. Your agent logic does not change.
03Which models and agent tools does it work with?
Any agent that calls a model through an Anthropic- or OpenAI-compatible API, including Claude Code, OpenRouter and DeepSeek. Receipts look the same whichever model or framework produced the action.
04Can Euthyne read my prompts and outputs?
No. Content is end-to-end encrypted by default and the keys stay with you. Receipts hold hashes of inputs and outputs, so they prove what happened without exposing what was said.
05What is Verana and why does Euthyne use it?
Verana is an open trust registry built on W3C verifiable credentials. Euthyne uses it to resolve which organization and which agent signed a receipt, so anyone can check an identity without having to trust Euthyne.
06How much does it cost?
Euthyne is in private beta. Pricing will depend on the number of agents, the volume of receipts and the assurance level you need (self-signed, countersigned or notarized), not on seats. Request access to join the beta.
Every action.
On the record.
Euthyne is in private beta with teams running agents in production. Request access and we will reach out.