Skip to main content
The prompts on this page work in any client connected to the Telivy MCP server. Each example shows the prompt and the tool calls the assistant typically makes. The exact tool sequence depends on the client and the model, and the results depend on the data in your Telivy account. The assistant runs every tool under your agency’s API User, so anything it creates or changes appears in Telivy as done by API User. For setup and the full list of tools, see the MCP server overview. Prompts work best when they name the client company or domain, say what you want back, and say whether the assistant may change anything. If a prompt asks for a change, read the tool call the client shows you before you approve it.

​
Start an external scan and summarize its findings

The assistant first calls get_external_scan_assessments to check whether a scan already exists for example.com, so it does not create a duplicate. If none exists, it calls add_external_scan_assessment with the organization name and domain, and your client asks you to approve that call. Telivy then runs the scan, which takes time. The assistant can check progress by calling get_external_scan_assessment again, but most clients do not wait in the background. If the scan is still running, ask again later in the same conversation, for example “Is the example.com scan finished yet?”. Once the scan is complete, the assistant calls get_external_scan_assessment_findings, groups the findings by severity, and can call get_external_scan_findings for the details and recommendation of an individual finding before it writes the summary.

​
List risk assessments with a failing scan status

The assistant calls get_risk_assessments, using its scanStatus filter where that narrows the list, then calls get_scan_status for each candidate to check device-level completion, and returns a table of the assessments that match. Both tools only read data, so clients that run read-only tools without asking complete this without approval prompts. Each get_scan_status call counts toward your rate limit, so with many assessments, narrow the prompt to a company or a time range.

​
Review the devices on an assessment and rescan one

The assistant finds the risk assessment for Example Dental with get_risk_assessments, calls get_all_deep_scan_devices to list its devices, and shows them with their last-seen times. It then calls rescan_deep_scan_target with the assessment ID and the ID of the device it matched by name, and your client asks you to approve the rescan. Check that the deviceId in the approval request belongs to FRONTDESK-01 before you approve it, because two devices can have similar names.

​
Compare breach data between two scans

The assistant finds the external scan for each domain with get_external_scan_assessments, calls get_breach_data for both, and compares the results: it reports the number of exposed accounts per domain and the breaches that appear in both. This uses only read-only tools. Breach data can include email addresses of your client’s staff, so share the answer only with people who should see that information.

​
Prepare a client-facing summary

The assistant reads the latest external scan and risk assessment for the company with get_external_scan_assessment and get_risk_assessment, their findings with get_external_scan_assessment_findings and get_device_findings, and where useful the description of a finding type with get_finding_details_by_slug, then writes the summary from that data. The instruction “Do not change anything in Telivy” makes the read-only intent explicit, and your client’s approval settings still stop any write call from running unless you approve it. The assistant cannot download the Telivy report itself, because report downloads are not available as tools; download the report from the portal if you want to attach it. Review the draft before you send it, and check every number and finding against the Telivy portal or the assessment report, because the assistant writes the wording itself and can misread or misstate the data.