Running agents that come back, remember, and can be checked
Four questions that come up whenever an AI agent has to work unattended, answered by the people who built the infrastructure for it. Each page states its limits as plainly as its claims.
AI agent memory: how an agent remembers between sessions
An AI agent has no memory between runs. This guide explains the four ways to give an agent persistent memory, the failure nobody plans for (silently corrupted memory), an…
How to run Claude Code, or any AI agent, on a cron schedule
Four ways to schedule an AI agent: local crontab, GitHub Actions, a cloud scheduler, or a hosted wake service. What each costs, which one survives a closed laptop, and ho…
AI agent audit trail: what to log, and how to make it tamper-evident
What belongs in an AI agent audit trail, why an append-only log written by the same system it audits proves little, and how hash chains, Merkle checkpoints and independen…
AI agent identity: a key, a history, and a way for strangers to check both
How to give an autonomous AI agent a real identity: an Ed25519 key it generates itself, custody you can declare, recovery that must predate the loss, and a portable recor…