Use cases

Machine traffic, judged one request at a time.

Three places where a provider-side boundary with no standing credential changes the outcome.

Agent & MCP tool calls

An AI agent calling a tool is a machine request like any other, except that it is generated at speed, in volume, and often by software you did not write. Trustplane sits in front of the tool surface and judges each call on its own merits: the agent proves where it runs, its attestation is matched against the anchor you declared, and the individual call carries a proof that is valid once. Proven agents are admitted and everything else is refused — per request, not per session, so a compromised caller cannot ride a credential it obtained earlier.

Built for AI agents and MCP servers.

Key-free service APIs

Internal services and partner clients normally get onboarded by handing out a credential, and from that moment the credential is the security model — copied into config, pipelines, and laptops. Trustplane replaces distribution with attestation: a workload that matches an anchor enrolls automatically, holds no secret, and produces a fresh proof for each request. There is no key to rotate, leak, or find in a repository, and revocation is one policy edit rather than a hunt through every place a secret was pasted.

Built for internal services and partner clients.

Evidence for every decision

Most authorization systems can tell you what is permitted, but struggle to tell you what actually happened. Because every request is judged at the boundary, every allow and every deny becomes a record with a reason code attached — which caller, which action, which route, and why the boundary decided the way it did. That turns access review and incident reconstruction into reading exportable evidence rather than inferring intent from scattered application logs.

Built for compliance and incident review.

Replay-proof financial workflows

In banking, the dangerous request is often a perfectly valid one sent twice. An attacker who captures a payment or transfer call from an automation workflow can resend it, and a bearer token will happily authorize it again. With Trustplane every request carries a proof that is valid once. The original call is verified and allowed; the replayed copy arrives with a proof that is already spent and is refused at the boundary, with a reason code in the audit record. The demo below shows this with an n8n workflow — the workflow itself is not rewritten.

Built for banking and payment automation. Demo scenario.

Built at Axelerant to solve our own problem.

Read the docs