Agents now move money, send email, update CRMs and negotiate with customers — but nobody independently checks whether they did what they claimed. IDRYN verifies agent outcomes where they actually happen, builds each agent's persistent behavior record, and turns it into the risk scores and evidence that approvals, disputes and insurance run on. FICO + Verisk, rebuilt for software that acts.
Independent by design · Not an insurer · Evidence, not opinions
Every day, AI agents issue refunds, move payments, write to customers and change records of account. When something goes wrong there is no independent record of what actually happened — only the agent's own logs, written by the agent itself.
An agent that says “refund sent” may have failed, hallucinated the transaction, or refunded the wrong customer. Today, the only witness is the agent.
Agents are redeployed, renamed and swapped between vendors. Their track record resets to zero each time — good agents can't prove quality, bad actors can't be screened out.
Enterprises can't set approval limits, platforms can't price risk, and insurers can't underwrite agent activity — because there is no trusted data on how agents actually behave.
IDRYN sits outside the agent, the platform and the insurer. We observe outcomes where they actually happen — in payment systems, inboxes and CRMs — and turn them into identity, scores and evidence everyone can rely on.
Every material claim an agent makes is checked against the system of record where the outcome lives.
Verified outcomes accumulate into a persistent, portable behavior record tied to a durable agent identity.
Records convert into a risk score and tiers that platforms, enterprises and insurers can price against.
Approvals, disputes, underwriting and claims are decided on signed evidence packs — not screenshots and log files.
Connect IDRYN to the systems agents act on: payment rails, Gmail/Outlook, Salesforce, HubSpot, helpdesks and internal tools — via API, webhooks or read-only connectors.
Each agent action registers as a claim: what was supposed to happen, for whom, for how much, and by when. Claims are timestamped before outcomes are known.
IDRYN independently queries the source system. Did the payment settle? Did the email actually send? Did the CRM record change — and to what?
Every claim resolves to Verified, Partial, Contradicted or Unverifiable — with the underlying evidence attached and discrepancies flagged in real time.
Resolved outcomes update the agent's behavior record and risk score, and package as evidence for approvals, disputes, audits and insurance claims.
Identity tells you who an agent claims to be. Evaluations tell you how its outputs read. IDRYN is the missing layer: independent proof of real-world outcomes, accumulated into risk data the whole economy can use.
| Eval / observability tools | Agent identity (KYA) | IDRYN | |
|---|---|---|---|
| What it checks | Output quality in traces | Who the agent is, at access time | Whether claimed business outcomes really happened |
| Source of truth | The agent's own logs | Identity tokens | Independent source systems — payments, mail, CRM |
| Persistent behavior record | — | — | Yes — portable across platforms & vendors |
| Risk score for counterparties | — | — | Yes — the IDRYN Score |
| Insurance underwriting & claims evidence | — | — | Yes — signed evidence packs |
Independent outcome verification for agent actions. Agents claim; IDRYN checks. Instead of trusting agent-generated logs, Verify confirms each material outcome in the system where it actually occurred — the same way a credit bureau confirms a payment with the lender, not with the borrower.
A durable identity and behavior file for every agent — the agent equivalent of a credit file. Records survive redeployments, vendor changes and renames, so a trustworthy agent carries its reputation with it, and a poor performer can't reset its past by changing its name.
A single, comparable risk score for any agent, computed from verified behavior — never from self-reported logs. Scores power practical decisions: approval thresholds, transaction limits, monitoring intensity, vendor selection and insurance pricing.
When money is disputed or a claim is filed, everyone asks the same question: what actually happened? Evidence packages the verified record — claims, source-system proof, timestamps and chain of custody — into a signed pack that enterprises, platforms and insurers can act on.
One scale for every agent, computed only from independently verified outcomes. New agents start unscored — never assumed safe.
Long, accurate, low-loss history. Suitable for high auto-approval limits and preferred insurance terms.
Reliable with occasional discrepancies. Normal limits, standard monitoring, insurable at standard rates.
Elevated contradictions or disputes. Reduced limits, full verification sampling, remediation plan expected.
Repeated contradicted claims or unresolved losses. Counterparties and insurers are advised to restrict or decline.
Differentiate quality agents with proof, not promises.
Let agents do real work — with controls that scale.
Underwrite the agent economy on data, not guesswork.
No. IDRYN is risk data infrastructure — the verification, records and scores that insurers, platforms and enterprises use. Like FICO for credit or Verisk for property risk, we supply the trusted data layer; underwriting and policies stay with licensed insurers.
Eval tools grade what an agent said, using the agent's own traces. IDRYN verifies what an agent did, using independent source systems — the payment processor, the mailbox, the CRM. And the result isn't a dashboard: it's a persistent record and score that counterparties and insurers can rely on.
Payments and refunds, sent communications, CRM and helpdesk record changes, and commerce actions like orders and bookings — anywhere a source system can be queried through an API, webhook or read-only connector. Claims on systems we can't reach are marked Unverifiable, never assumed true.
The record belongs to the agent's operator. IDRYN maintains it independently — that's what makes it credible — and operators control which counterparties can look it up, the same way businesses control their credit file access.
We onboard a small number of platforms, enterprises and insurers to define the first verification schemas, score bands and evidence formats together. Partners get early API access and direct input into the standard. Apply below — we reply to every serious application personally.
We're onboarding a small cohort of design partners — agent platforms, enterprises deploying agents at scale, and insurers preparing agent coverage. Early partners shape the verification schemas, score bands and evidence formats everyone else will use.
Apply as a Design Partner →partners@idryn.com · We reply personally · No deck required