# Monitoring Handoff (template)

> **TEMPLATE. Manually prepared service deliverable.**
>
> - Use it when a customer moves from observation to ongoing monitoring.
> - Monitoring does not enable blocking. A customer can stay in monitoring indefinitely.
> - Bracketed values are agreed during scoping. Nothing here is a published service level.

| | |
|---|---|
| Customer | [name] |
| Provider | [name] |
| Effective from | [date] |
| Coverage statement | [version] |
| Enforcement posture | **monitor** (Behavry tenant setting). Changing it requires the protection change process in section 6. |

## 1. Signals to watch

Agree each signal explicitly. Only signals that a connected source can produce belong here.

| Signal | Source | How it appears | Severity | Review cadence |
|---|---|---|---|---|
| A policy would have blocked or held an action on a governed path | Governed path in monitor posture | `monitor_would_block` event, with the observed verdict on the decision record | [warning] | [daily / weekly] |
| DLP finding on AI input | Browser extension | Browser event with DLP findings | [warning / critical] | [ ] |
| New AI app grant with sensitive scopes | IdP connector | New `idp_oauth_grants` row | [info / warning] | [ ] |
| Citizen app ungoverned for 30 days | Citizen-code connector | `citizen_app_ungoverned_30d` alert | [warning] | [ ] |
| New MCP server configuration on a device | Desktop agent | Exposure finding | [warning] | [ ] |
| [custom] | | | | |

## 2. Routing

Alerts are delivered to [SIEM: Splunk / Sentinel / QRadar / Chronicle / syslog / webhook] and/or [Slack / email]. Connection is confirmed per customer.

Alert statuses are `open`, `acknowledged` and `resolved`. The product records who resolved an alert and when. If you need a separate record of who acknowledged an alert, keep it in your ticketing system.

## 3. Who does what

| Activity | Behavry | Provider | Customer | Notes |
|---|---|---|---|---|
| Keep the platform running and the connectors' runtime healthy | **R/A** | I | I | |
| Keep agreed sources connected; report gaps | C | **R** | **A** (grants access) | A gap changes the coverage statement |
| Review alerts within the agreed cadence | | **R/A** | I | |
| Investigate and write up an alert | | **R** | C | |
| Recommend an action | | **R** | **A** (accepts or declines) | A recommendation is not an approval |
| Approve a change in a customer system | | C | **R/A** | Named approver: [name, role] |
| Make the change | | R only if requested | **A** | Under the customer's change process |
| Notify affected users | | C | **R/A** | |
| Produce the evidence summary | | **R** | **A** (receives) | Cadence: [ ] |
| Access raw evidence and exports | | R | R | Audit export JSON/CSV/CEF/LEEF; evidence package where available |

R = responsible, A = accountable/approves, C = consulted, I = informed.

## 4. Escalation contacts

| Situation | Provider contact | Customer contact | Target response |
|---|---|---|---|
| Critical alert | [ ] | [ ] | [agreed in scoping] |
| Suspected incident | [ ] | [ ] | [the customer's incident process applies] |
| Source disconnected | [ ] | [ ] | [ ] |

## 5. What monitoring does not do

- It does not block, hold, revoke, disable or change anything automatically.
- It covers only the connected sources in the coverage statement.
- The safety checks listed in the coverage statement still apply.

## 6. If the customer later wants protection

Protection is optional and is enabled action by action.

1. The provider proposes specific controls: the action, the path, and the outcome (allow, hold for a named human, or deny). Include the evidence from monitoring that supports each one.
2. The customer's named approver accepts the proposal in writing, through the customer's change process.
3. The control is enabled for the agreed scope. The rest stays in monitoring.
4. Rollback: [how, who, and how quickly].

**Note on scope.** Posture is set per tenant, and a child tenant does not inherit it from its parent. Confirm each tenant's posture after any change.
