Features

One identity layer for every agent.

Oathera gives every AI agent an identity bound to the machine it runs on, with a named person who answers for it, and checked every time it acts. These are its capabilities, grouped by what they do for you.

One identity layer with five capability groups A central agent identity marked with the Oathera logo, with five capability groups around it: identity, approval, proof on every call, safe autonomy, and records. Each lights up in turn. ONE IDENTITY LAYER Identity Approval Proof Autonomy Records agent identity
An identity that can't be copied

Bound to the machine it runs on.

An anchor key for every agent

Each agent gets one key, created and kept on its own host. The agent's identity is the key digest of that key, so logs show the agent, not a shared key. Moving an agent to a new anchor key does not mean enrolling it again.

Host binding

The host is recorded at enrollment and compared at every token request. Enforce refuses a mismatch; warn, the default for testing and design partners, issues the token and raises a high-severity alert. A reboot is not a new host.

Platform attestation

On Google Cloud, the platform signs an instance identity statement bound to the agent's key, and a setting makes it mandatory. AWS and Azure documents are checked in part and don't reach verified.

Hardware or KMS key custody

The anchor key can be held in a key manager, hardware module, TPM or secure enclave, and the private key never leaves it.

Protected core

The enforcement logic loads from a signed native library, checked before it loads and while it runs. A changed library stops the MCP.

A person answers for every agent

Approved by a named person, stopped on demand.

Enrollment approval through your identity provider

A principal signs in through your identity provider and approves each agent. The approval is part of the agent's enrollment record.

A named, verified sponsor

Record the person or organization who answers for each agent, verified by your own proofing sources.

Discovered agents

The console lists agents Oathera Agent Enforcement refused for having no identity, so you can enroll or reject each one.

Revoke an agent

Revoke an agent from the Oathera Identity Service and its next token request and its next boundary decision are refused. A token already issued stays valid until it expires.

Deprovision a person

Stop every agent a principal approved, without touching the person's own sign-in. When the identity provider reports a departure, through SCIM or Shared Signals, the agents stop without an administrator's call.

Kill switch with a way back

Suspend an agent, then reinstate it after its host is checked again, with reduced access first.

Handover

When a principal leaves, their agents are suspended and another principal of the same sponsor approves each one again within a grace period.

Verifiable credentials

The credential service records who approved what as verifiable credentials, with suspend, reinstate and revoke status.

A proof on every call

A stolen token authorizes nothing.

Short identity tokens

Identity tokens last five minutes at most. Each one names a new session key that lives only in memory, so a copied token is useless without it. A token can be minted for one service and be useless against any other, with a shorter life for privileged actions.

A proof on every call

Before each call the agent signs a single-use operation proof over the operation and its arguments. Replay it and it is refused; change an argument and it is refused.

Gateway checks and boundary decisions

Oathera Agent Enforcement checks the token and the proof, then asks the Oathera Identity Service for a signed, single-use decision. No valid decision, no call.

Gateway assertions

Your service gets a short, single-use statement that one request passed, made for that service only, in place of the agent's token.

Tool and argument limits

Oathera Agent Enforcement refuses unknown tools and arguments outside a tool's allow-list. A default-deny setting also refuses any tool you haven't configured.

Your own tools, in any language

Oathera Agent Enforcement verifies proofs for any tool you configure, and a gateway proxy sits in front of MCP servers written in any language, over stdio or Streamable HTTP.

Clearance

An agent holds the lower of the clearance requested for it and what its evidence supports. An agent with stronger evidence reaches operations that need more than the baseline.

Content trust

Arguments that come from the model are labelled as agent content and can't set protected security fields. Oathera doesn't inspect content for prompt injection.

Safe autonomy

Agents that act on their own, inside limits.

Task context, delegation and approvals

A signed task context bounds what an agent is doing. Delegation to a sub-agent can only narrow its authority. A person's single-use approval covers one exact operation.

Evidence grades

Evidence is graded as self-reported, corroborated or verified, so self-reported evidence never meets a policy that needs attestation.

Risk tier from the NHI questions

When an agent answers the OWASP non-human identity questions, the Oathera Identity Service derives a risk tier from 1 to 4 and records the findings. The answers are self-reported.

Agent-to-agent verification

One agent checks another agent's identity before it trusts it.

Records you can trust

Every decision, in order.

Hash-chained audit log

The Oathera Identity Service links each security decision into a hash chain and replaces credential-like values with a correlation ID. The log is sealed in signed batches, verified as a chain, and exported exactly once to your SIEM or a write-once store.

Faster revocation status

Signed revocation snapshots are refreshed faster for agents with a more dangerous risk tier.

The Oathera console

The console has its own sign-in, admin and viewer roles, and a step-up code to create a team. It shows live decision, posture and agent-risk views, fed by the MCPs.

Admin roles in the Oathera Identity Service

Role checks with separation of duties, and a fresh sign-in for high-risk actions.

Shared Signals

The Oathera Identity Service publishes agent revoked, suspended and host-changed events over SSF and CAEP, and accepts an account-disabled event from your identity provider.

Fits what you run

Beside your identity provider, not instead of it.

Federation with Microsoft Entra

Exchange an agent's Oathera identity for a token Microsoft Entra accepts.

SPIFFE evidence at enrollment

Accept a SPIRE-issued workload identity as extra evidence when an agent enrolls. The agent's identity stays its anchor key.

Rollout stages

Watch first, enforce for a few agents, then enforce everywhere.

Tell us what to build next.

Design partners get a say in the roadmap. Run Oathera in front of one real tool with real agents, and tell us what breaks.