Technical deep-dives on agent identity, behavioral attestation, and trust infrastructure.
November 6, 2026
Some sites ask scripts for a contact in the User-Agent header. Give them a contact for the agent, never the human's login email. The options, what Wikimedia and the SEC ask for, what six sites' filters actually let through, and what to do about Claude Code's email leak.
November 5, 2026
Check three things in order: who the other agent is (a signature), what it is allowed to do (scopes), and what it has done before (records and track record). What each check can and cannot tell you, including the gap in our own product.
November 5, 2026
In a secrets manager the agent reaches with a short-lived identity, and ideally never in the model's context at all. Cloud secret managers, HashiCorp Vault, 1Password, and what AgentLair's vault does and doesn't do.
November 5, 2026
Three ways to give an agent a mailbox it can send from and read, which email APIs are built for agents (AgentMail, Resend, Nylas, AgentLair), and the commands to try one in a minute.
November 5, 2026
Log every run, model call, tool call and approval with shared IDs, send it somewhere the agent cannot edit, and make entries tamper-evident. OpenTelemetry for the traces, a signed log for the record, and a check you can run yourself.
November 5, 2026
An agent proves who it is by signing with a key the other side can check: cloud workload identity, SPIFFE, Web Bot Auth, or a signed JWT like AgentLair's AAT. What each proves, what none of them prove, and a token you can verify yourself.