coming soon
film
Teach it once
The process is taught once. Every AI and every person then runs it the same way.
learn: The ideas behind governed AI work.
Guides and a working glossary for governed AI agent workflows, from the team behind ConvOps. Every idea comes with a picture of how it moves: plain words first, the engine second.
12 guides · 25 terms · films coming soon
The whole systemAutonomous systemAn AI loop that starts itself, follows a process, and leaves a record.
the guides
Guides for running AI agents under a process.
Long reads with a picture for every idea, from making Claude Code follow a process to EU AI Act oversight. Start with the one closest to your question.
the category guide
How to govern autonomous AI systemsProcess as structure, gates where judgment is needed, two names in the audit. Read the guide the mechanics guide
Agentic workflows, explainedSteps, gates, detours and loops: how a process becomes something an AI can run. Read the guide the memory guide
Why AI agents forget, and the memory that makes run 40 beat run 1Why AI sessions forget, and the capture habit that makes run 40 beat run 1. Read the guide claude code
How to make Claude Code follow a processSteps as a skill, hard stops as hooks, and an engine for gates and sign-offs. Read the guide claude code
CLAUDE.md too long? What to keep and what to moveUnder 200 lines. Facts stay; procedures, folder rules and hard rules move out. Read the guide claude code
How to run Claude Code unattended on Kubernetesclaude -p in a locked-down Job, then a process that knows where to stop. Read the guide governance
Human approval gates for AI agentsA stop the system checks, not a request the model weighs. Read the guide the definition
What is graph engineering? A guide for AI agent teamsThe term's three meanings, with sources, and the process-graph sense ConvOps uses. Read the guide the audit guide
AI agent audit trail: what to record, and whyWhat an agent audit trail must record, and why the human identity has to come from the credential. Read the guide the regulation guide
EU AI Act human oversight for AI agentsArticle 14 in plain words, the post-omnibus dates, and how oversight maps to agent workflows. Read the guide the comparison guide
Skills vs MCP vs workflows for AI agentsSkills package know-how, MCP gives reach, workflows set order and stops. Fair definitions, side by side. Read the guide the team memory guide
Shared memory for AI agent teamsWhat Claude Code shares across a team, what it keeps on one machine, and how to add a shared brain. Read the guide Weighing tools? ConvOps compared with Dynamic Workflows, Archon, LangGraph and more
the glossary
Every term, and how they connect.
Hover or tap a term to see what it touches. Colour tells you the kind of idea. Each one opens to a full definition.
The process
Agentic workflowA process the AI runs step by step, checked by an engine.MCP (Model Context Protocol)The open protocol that plugs ConvOps into your AI tool.Workflow policyA rule delivered inside the step where it applies.Graph engineeringSmall workflows behind one door, walked one step at a time.Workflow routerOne door for new work. A decision picks the exit.Sub-workflowA detour on the same task. The parent waits, then carries on.Fix loopFailed work goes back through a fix and comes round again.Workflow fragmentShared steps written once, embedded in many workflows.MCP serverExposes tools and data to any AI client that speaks MCP.Human decisions
Workflow gateA condition the engine checks. Until it is met, the run waits.Human-in-the-loopThe process stops at named moments for a person to decide.Dual attributionEvery change records the agent and the signed-in person, separately.Governed AI agent workflowA process agents run step by step, with stops the engine checks.Row-level security (multi-tenant)The database itself keeps each tenant's rows apart.Running it
Attended executionYou drive the AI. The engine supplies the steps.Unattended executionNobody at the keyboard. People join only at gates.Work poolA live saved query that hands the loop one task, or none.Child tasksSplit the work. Same order starts together. The parent waits.MCP clientThe part of your AI tool that talks to MCP servers.Agent executorWhich agent runs the work, and where.Headless agentAn AI agent run with no interface and nobody watching.the films
Start with the films.
Short films and a course that show the same ideas running. They are being finished now.
coming soon
film
Teach it once: the brain
A decision made in one AI tool, still there when another tool opens a new chat.
coming soon
course · 9 lessons
Govern your repo
Approvals on every step first. Then let the work run on its own as trust grows.
your next step