AGENT INVENTORY BUILDER
Name your agents. Get the inventory an auditor asks for.
An auditor, an insurer or the board asks which AI agents act in your name, who owns each one, what it can reach, what it can do, who approves, and how you stop it, and at most firms the honest answer is a list nobody has written. Name up to six agents, flows or automations, answer eight short questions about each, and this page writes the inventory: one row per agent with its owner, reach, actions, approval step, identity, stop path and last review; a finding on every row with the control that would close it (a named Microsoft capability with the date it was checked, or the setting in your own tools); the three fixes to make first; and the rules that apply. Download it as CSV for your register, or as PDF, Markdown or JSON. It works for Microsoft 365, Google Workspace and mixed stacks. No email is needed.
UP TO SIX AGENTS
Name each agent. Answer eight questions about it. The inventory writes itself.
Three questions about the organization, then one block per agent: a flow, a Copilot agent, a custom GPT, a vendor's built-in agent, a script. Use the names your team uses. The inventory is written on this page the moment you finish and gets an address of its own. Do not enter credentials, keys or customer records.
Keep a copy in your inbox.
Enter a work email, company and role to receive this inventory as text with its address. This records an SIT inquiry for follow-up, and the engineer reads it before any call. Marketing emails require a separate choice.
YOUR REGISTER
What this browser has worked out so far.
Kept in this browser only; nothing on this list is stored on our server. Each line opens its own address, so you can come back to it, send it to the owner, or clear it. The procurement pack binds every line into one PDF, rebuilt from the addresses when you press the button, with a cover and a contents list.
WHAT IT IS
What is an agent inventory?
One list of every AI agent, flow and automation that reads or acts in your name, with the same seven columns for each: owner, what it may read, what it may do, who approves, what it signs in as, who can stop it, and when it was last reviewed.
The columns are the point. An agent with a name and no owner is a liability with a label. An agent that can pay or change permissions with nobody approving is a finding in any audit, whatever the vendor calls it. The builder writes the row, then says which column is a gap and what closes it.
THE INSIGHT
Why does every agent need its own identity and a stop path?
Because an agent that runs on a person's login can reach everything that person can and the log cannot tell the two apart, and an agent nobody has tried to stop will be stopped for the first time during the incident.
Microsoft Entra Agent ID is generally available to every Microsoft Entra customer and gives each AI agent its own identity with owners and sponsors, Conditional Access templates for high-risk, autonomous and on-behalf-of agents, and sponsor lifecycle workflows; extending Entra security features to agents needs a Microsoft Agent 365 license, and the admin center creation wizard and agent user accounts remain in preview. (Microsoft Learn, checked Nov 4, 2026)
Write the identity and the stop path down and the first incident takes minutes. Leave them blank and it takes the weekend, with the agent still running while someone looks for the password.
HOW IT WORKS
How does the builder write the inventory?
From a checks table, not a model: each of the eight answers about an agent sets a finding to in place, partly or gap, and picks the control for your platform, so every row traces to an answer and, on Microsoft 365, to a dated register entry.
- You name up to six agents and answer eight questions about each; nothing is sent anywhere.
- The page writes the table, a finding for every column, the three fixes to make first, and the rules that apply, with the control that closes each finding.
- You download it as CSV for your register, or as PDF, Markdown or JSON, and it has its own address.
- The ten-hour foundation checks each row against the tenant; until then the inventory is self-reported, and says so.
ANY STACK
What if our agents are not on Microsoft 365?
Then every control names the service account, the OAuth scope or the automation platform's own setting instead of a Microsoft capability, and the columns stay the same, because an auditor asks the same seven questions whatever the agent runs on.
Microsoft 365 shops get Agent 365, Entra Agent ID and Defender named with the date each was checked, and a note where the capability is still in preview. Everyone else gets the dedicated account, the scoped consent and the tested off switch.
YOUR DATA
What happens to the answers?
They are rendered on the page and encoded in the inventory's address, and nothing is stored unless you ask for the email, in which case the lead is recorded through the same intake as the contact form.
The only free text is the agents' names and your organization's name. Use the names your team uses; do not enter credentials, keys or customer records.
THE MONEY
What is free and what is paid?
The inventory is free. Checking each row against your tenant, marking it tested instead of self-reported, and closing the first three fixes is the ten-hour foundation, and anything built after that is scoped from it.
No price appears on this site; the engagement is quoted after a scoping call, assessment separate from implementation.
AFTER TWO WEEKS
What will be working after two weeks?
- You hold a one-page inventory of up to six agents with the same seven columns an auditor asks about, and a status on each row.
- Every finding names the control that closes it, with the date the Microsoft fact was checked where one applies.
- The three fixes to make first are ordered by consequence: money, access and code changes without a person come before an overdue review.
- The inventory downloads as CSV for your register and has its own address, so the owner can reopen it next quarter and change one row.
HONEST ABOUT LIMITS
What this does not do yet
It does not know what your tenant knows. Agents built with Microsoft Foundry, Copilot Studio and Agent Builder are integrated with Microsoft Agent 365 automatically; agents built elsewhere join through registry sync or the Agent 365 SDK. (Microsoft Learn, checked Oct 31, 2026) That registry is the list the inventory should be checked against; the ten-hour foundation reads it and marks each row found or missing.
It does not discover agents: it lists the ones you name, and the tenant may know about more.
It does not test a control: a stop path marked "tested" is your word until read-only discovery checks it.
THE ONE THING TO REMEMBER
An agent nobody owns, nobody approves and nobody can stop is not automation; it is an incident with a start date.
Said another way: seven columns per agent, or no agent.
Said a third way: the list is free; checking it against the tenant is ten hours.
DIRECT ANSWERS
Questions before you decide.
Do I need an email to see the inventory?
No. The inventory renders on the page and has its own address. The email is only for keeping a copy and it records a lead through the normal intake.
Is the inventory written by an AI?
No. It comes from a checks table, so every finding traces to the answer that set it and to the control named beside it. A model may tailor wording in a later version, under the same rules.
What counts as an agent?
Anything that reads or acts in your name without a person typing each step: a Copilot agent, a flow, a custom GPT with actions, an automation platform workflow, a vendor's built-in agent, or a script your team wrote.
Can I list more than six?
Six at a time. Each inventory gets its own address, so run it again for the next six and keep both in the register. Six is already more than most teams can bring under control in a quarter.
How does it relate to the policy builder?
The policy says an inventory exists and is reviewed every quarter. This page is that inventory, with the same owner, approval and identity rules, so the two documents agree.
NEXT STEP
Start with a boundary you can explain.
Choose a focused check, or speak with the team about the workflow you need to improve.