Agent authority verification

Every agent action, traced back to a person.

Ipseiti credentials AI agents and the mandates behind them, checks that authority before an action executes, and signs the result into a record no one can quietly rewrite.

PrincipalDana OrtizIdentity proofed by a partner
MandateBuy compute ≤ $5,000/moExpires 31 Mar 2027
Agentprocurement-agent-07Holds agent credential
ActionPlace order, $1,240.00Within scope and limit
RecordAuthorized · signedLog entry #48213
01Position

IAM manages access. Guardrails manage safety. Ipseiti owns authority.

When an agent spends money, signs a customer or moves data, someone will eventually ask who allowed it. Existing controls weren't built to answer that.

QuestionIAMGuardrailsIpseiti
Can this agent reach the system?YesNoRelies on IAM
Is this action safe or on-policy?NoYesWithin mandate scope
Which person authorized this exact action?NoNoYes
Can you prove it later, in a dispute?Access logsEvent logsSigned, tamper-evident
02Method

Issue. Verify. Prove.

ISSUE

Credential the agent and its mandate

Ipseiti issues an agent credential and a mandate credential stating what the agent may do, for whom, up to what limits and until when. Partners handle identity proofing of the human principal.

VERIFY

Check authority at the moment of action

Before a payment, contract or data transfer executes, the receiving service asks Ipseiti whether this agent holds a valid, unrevoked mandate for this exact action.

PROVE

Keep a record that holds up

Each decision is signed and appended to a hash-chained log. If the action is challenged, you can show who authorized it, on what terms, and when.

03Non-repudiation

An append-only record of who authorized what.

Each entry carries the hash of the one before it. Edit or delete an entry and the chain breaks where anyone can see it.

verification logexample
#14 Jan 2027, UTCAgentActionDecisionEntry hash
4821117:01:52support-agent-02refund.issue 89.00 USDauthorizeda44f…7b10
prev 5c0d…e2a1
4821217:02:03procurement-agent-07order.create 6,800.00 USDover limitb71e…09ad
prev a44f…7b10
4821317:02:11procurement-agent-07order.create 1,240.00 USDauthorized9f3a…c41e
prev b71e…09ad
4821417:02:40data-agent-11export.customers EUmandate revoked0d72…5b98
prev 9f3a…c41e

Illustrative data. The record format may change before launch.

Signed decisions

Every result is signed by Ipseiti, so neither party can later deny what was checked.

Scoped mandates

Limit agents by action type, amount, counterparty and time window.

Revocation

Withdraw a mandate or agent credential and the next check reflects it.

Hardware-held root keys

Root and intermediate signing keys stay inside a hardware security module.

Partner identity proofing

Established identity providers verify the person behind each mandate.

Pay per verification

No seats or minimums. Pay for each check, machine-to-machine payments included.

04Integrate

One check, wherever your agents act.

REST

Verify API

Call one endpoint before executing an agent's request and get a signed decision in the response.

MCP

MCP server

Agent frameworks present and check credentials through a native MCP tool.

Edge

Low-latency checks

Verification runs at the network edge, close to the services your agents call.

05Questions

Agent authority, answered.

What is Ipseiti?

Ipseiti is an agent authority verification service. It issues verifiable credentials to AI agents and to the mandates that empower them, checks that authority before an agent's action executes, and signs each decision into a tamper-evident record.

How is Ipseiti different from IAM or AI guardrails?

Identity and access management (IAM) controls which systems an agent can reach, and guardrails judge whether an action is safe or on-policy. Neither records which person authorized a specific action. Ipseiti binds each agent action to a mandate and a verified human principal, and keeps proof of it.

What is a mandate credential?

A mandate credential is a signed statement of what an AI agent may do on someone's behalf: the type of action, spending or volume limits, permitted counterparties and an expiry date. Services check the mandate before accepting the agent's request.

How does Ipseiti provide non-repudiation for AI agent actions?

Every verification result is signed by Ipseiti and appended to a hash-chained log, where each entry includes the hash of the one before it. If an action is disputed, the record shows who authorized it, on what terms and when, and any later edit to the log is detectable.

How do developers integrate Ipseiti?

Services call a single REST verify endpoint before executing an agent's request, or use Ipseiti's MCP server so agent frameworks can present and check credentials natively. Pricing is per verification, including machine-to-machine micropayments.

When will Ipseiti be available?

Ipseiti opens as a private beta in January 2027 for teams building or accepting autonomous AI agents. You can request access on this page.

● Private beta · January 2027

Put a name behind every agent action.

We're opening Ipseiti to a small group of teams building or accepting autonomous agents. Leave your details and we'll be in touch as we launch.