# Coverage and Limitations Statement (template)

> **TEMPLATE. Manually prepared service deliverable, not a product export.**
>
> - Complete it with the customer before the observation window starts.
> - Re-issue it whenever a source is connected or removed.
> - Bracketed text is guidance. Replace or delete it.

| | |
|---|---|
| Customer | [Customer legal name] |
| Provider | [MSP/MSSP name] |
| Platform operator | Behavry |
| Statement version and date | [v1.0, YYYY-MM-DD] |
| Service stage | [Observation / Monitoring] |
| Observation window | [start] to [end] |

## 1. Connected sources

List only the sources that are actually connected and syncing. A source that is "available" but not connected gives no coverage.

| Source | Connector | Access granted (scopes or permissions) | Scope (users, groups, devices) | Connected on | What it can show | What it cannot show |
|---|---|---|---|---|---|---|
| [e.g. Microsoft 365 admin API] | [microsoft_365] | [admin consent, Graph application permissions as configured] | [whole tenant] | [date] | AI feature licensing and enablement, enabled user counts | Prompts, content, per-user actions |
| [e.g. Entra ID / Okta / Google Workspace OAuth apps] | [azure_ad / okta / google_workspace] | [read-only directory or app scopes as configured] | [tenant] | [date] | Third-party apps, granted scopes, principals | Whether an app was used; sign-in or audit logs |
| [e.g. Managed browser extension] | browser extension | Managed deployment via [MDM] | [n of N managed browsers] | [date] | AI site visits, paste and input events, DLP findings | Unmanaged browsers, personal devices, native apps |
| [e.g. Desktop agent] | Warden | Device enrolment | [n devices] | [date] | LLM API call metadata, MCP configuration exposure | Devices without the agent |
| [e.g. Citizen-code platform] | [replit / v0 / vercel / netlify] | [platform API token, read] | [workspace] | [date] | Deployed apps, creator metadata | Platforms not connected |
| [e.g. Governed MCP or API path] | proxy | Agent points its tool configuration at Behavry | [named agents] | [date] | Individual tool calls on that path, with the policy decision | Calls made by any other path |

## 2. Three levels of visibility

- **Service detected.** An AI service, licence, app grant or deployment exists. This does not establish use.
- **Usage observed.** A source shows people or devices using a service (browser events, desktop agent metadata).
- **Action recorded.** An individual action was recorded on a configured, governed path, with the policy decision.

Discovery-only connectors (admin APIs, OAuth grants, citizen-code platforms) stop at *service detected*. They do not capture tool calls and do not produce action-level Decision Traces.

## 3. Not covered

[Customise. Keep the defaults that apply.]

- Network, DNS, proxy or firewall traffic. These are not a Behavry discovery source.
- Identity-provider sign-in or audit logs.
- Personal devices, unmanaged browsers, and devices without the desktop agent.
- SaaS platforms and citizen-code platforms that are not connected.
- AI used through accounts or tenants the customer does not control.
- The content of prompts and responses, except DLP pattern findings where the browser extension is deployed.

## 4. What observation and monitoring do not do

- They **do not block, hold, or change** any action, account, grant, or setting.
- A finding or an alert is reviewed by a person. Any change is decided by the customer's named approver and made through the customer's change process.
- Some safety checks apply at every stage, so observation is not a promise that nothing is ever refused:
  - agent authentication;
  - session and tenant-ownership checks on governed paths;
  - unregistered MCP servers on a governed path;
  - an administrator's emergency lockdown.
- In monitoring posture, a policy-engine outage does not block traffic. Actions continue and are recorded where the path allows.

## 5. Data handling

[Where findings and evidence are stored. Who at the provider and at Behavry can access them. Retention as agreed. Reference the customer's data processing terms.]

## 6. Responsibilities

The table below is a summary. The full split is in the *Monitoring handoff*.

| | Behavry | Provider | Customer |
|---|---|---|---|
| Operate the platform and connectors' runtime | Yes | | |
| Connect and maintain the agreed sources | Supports | Yes | Grants access |
| Review findings and alerts | | Yes | |
| Recommend changes | | Yes | |
| Approve changes | | | Yes |
| Make changes in customer systems | | Only if the customer requests it, under their change process | Yes |
| Access evidence | | Yes | Yes |

## 7. Acknowledgement

| Role | Name | Date |
|---|---|---|
| Customer | | |
| Provider | | |
