Getting started with AgentProof

Connect Microsoft

Open your workspace and click Connect Microsoft. AgentProof will request read-only permissions against your Power Platform environments. After sign-in, AgentProof discovers the agents your account can see and you pick one to assess.

Add a non-Microsoft agent

If your agent is not in Microsoft (a website chatbot, an internal agent, an OpenAI-style custom GPT, a workflow automation agent), use Add a non-Microsoft agent. You will describe it in plain English; AgentProof will label the readiness report as User-declared manual agent profile with the evidence basis you supply.

Try demo data

Demo mode (when enabled) lets you assess two sample agents without connecting your own tenant. Demo data is isolated and never enters a real workspace's report history.

Generate a report

Once questions are answered, click Generate readiness report. The report covers the score, risk view, control priorities, evidence basis, and a buyer-safe disclaimer. Print or save to PDF directly from the browser; AgentProof handles its own page furniture.

What scores mean

The 0 – 100 readiness score is a band-aware summary. The band (Ready with controls, Needs review, Not ready, Insufficient evidence) and risk posture are the headlines; the score is just the single number you can paste into a status update.

What to do after a report

The "Next best action" card identifies one improvement that would most lift the score. The Top 5 priorities are the focused list of fixes. Re-score after applying any controls; AgentProof keeps the prior reports as history.

Common issues

• No environments found. The signed-in account may lack Power Platform access, or admin consent may be required for the tenant.
• Manual agent is missing. Manual data persists per workspace; switching workspaces shows a different set.
• Re-score lost my answers. Re-score resets answers by design so the new report is fresh. The prior report remains in history.