learn
The discipline, written down.
Governing autonomous AI systems is becoming a practice with its own vocabulary and its own patterns. These guides and definitions are how we think about it: concept first, mechanism second, and every claim about the product held to the same bar as the product pages.
the guides
the category guideHow to govern autonomous AI systemsA practical framework for governing AI agents that work without supervision: encode the process as structure, place gates where judgment is required, deliver rules at the moment of action, and keep two identities in the audit.9 min read →the mechanics guideAgentic workflows, explainedWhat agentic workflows actually are, how the pipeline style differs from the conversation-first style, and how steps, gates, decisions, and loops fit together into a process an AI can run repeatedly.8 min read →the memory guideMemory for AI agents: why run 40 beats run 1Why AI sessions forget, what agent memory actually is (and is not), the WHAT/WHERE split that makes recall useful, and the capture discipline that makes a loop compound instead of resetting.7 min read →
the glossary
Definition-first entries for the vocabulary of the category. Each answers its question in the first sentence.
Autonomous systemAn autonomous system is an AI-driven loop that starts itself, selects its own work from the current state of the world, executes a defined process, pauses where humans are required, and leaves an auditable record.Agentic workflowAn agentic workflow is a defined, machine-evaluated process that AI agents execute step by step, with instructions delivered per step and conditions that control advancement.Workflow gateA workflow gate is a condition attached to a step that an engine evaluates on every attempt to advance; until the condition is met, the workflow refuses to move.Human-in-the-loopHuman-in-the-loop is a design pattern where an automated process pauses at defined points for a human decision, rather than the human supervising continuously or not at all.Attended executionAttended execution is the mode where a human drives the AI session while a workflow engine supplies the process: the agent asks for the current step, does the work with the human present, and advances.Unattended executionUnattended execution is the mode where a scheduled or triggered run executes with nobody at the keyboard: a dispatcher hands the work to an execution backend, and humans participate only at gates.Work poolA work pool is a saved, typed query over work items that an autonomous loop consults to pull its next task; it is resolved live at each call and returns one item or nothing.Agent memoryAgent memory is persistent, retrievable knowledge that carries across AI sessions: what past runs learned, what was decided and why, and where things live, recalled before acting rather than rediscovered.MCP (Model Context Protocol)MCP is an open protocol that lets AI clients connect to external servers that provide tools, resources, and context, so the same server works from any client that speaks the protocol.Dual attributionDual attribution is an audit design that records two separate identities on every action: the agent's self-declared label, and the human derived server-side from the verified credential.Workflow policyA workflow policy is a reusable governance rule attached to workflows and delivered to the agent inside the step it is currently executing, so the rule is in context at the moment it applies.