Resources
Run AI agent work to a clear finish.
Plain guides for goals, agent runs, human decisions, cost, evidence, and accepted results.
Announcement
Announcing Workspace Memory
Agents start every task from zero. Workspace Memory keeps what the work proved, with a source on every line, a human who can overrule it, and one store that Copilot, board agents, and governed Claude Code and Codex sessions all read.
Workspace Memory guide
What should your AI agents remember?
Useful agent memory is not a transcript. It is a small set of cited decisions, constraints, facts, procedures, and risks that people can inspect and change.
Foundation guide
What is AI agent operations?
A plain-language guide to the work needed after an AI agent can produce an output.
Governance guide
Human approval workflows for AI agents
How to add a real human decision to an agent workflow without losing the action or its evidence.
Measurement guide
AI agent cost tracking for accepted work
A measurement model that connects AI agent spend to workflow quality and accepted results.
Illustrative internal example
From request to accepted result: an Aestus walkthrough
A checked-in, fictional Aestus example that shows the difference between a completed run and an accepted result.