@agentguard-run/trace

Signed tamper-evident audit trails and DAG trust chains for AI agent execution. Ed25519 receipts chain across turns and agents. Cannot be altered post-execution, even by the infrastructure operator.

latest

npm

Version: 0.1.0

Version published: 4 months ago

Maintainers: 1

Created: 4 months ago

Source

@agentguard-run/trace

Signed, tamper-evident audit trails and DAG trust chains for AI agent execution.

Every agent action is Ed25519-signed at execution time. Receipts chain across turns and agents via directed acyclic graph (DAG) attestation. Cannot be altered post-execution, even by the infrastructure operator.

Status

In production. AgentGuard Trace secures autonomous decisions across 6 production ML models, 11M+ sensor readings, and 320 site-years of data in a live commercial agricultural AI deployment.

Full SDK available under private preview. Contact invest@agentguard.run for evaluation access, OEM licensing, or strategic partnership inquiries.

How it works

Agent A ──► [Ed25519-signed receipt] ──► Agent B ──► [Ed25519-signed receipt]
                                              |
                                              v
                                       DAG attestation graph
                                       (every hop independently verifiable)

No operator can forge or delete a valid Ed25519 signature without the private key, which never leaves the customer runtime.

Patent notice

See PATENTS.md. Protected by 5 U.S. patent-pending applications filed February and May 2026, covering the DAG attestation architecture, signed receipt chaining, and multi-agent trust chain.