It does the looking up, and shows you each step

Give it a goal rather than a query. It plans, calls tools against the control graph, and reports what it found. Each step is streamed as it happens, so you watch it work instead of waiting on a spinner and then trusting the summary.

The reason this is worth having rather than asking a chatbot: the agent has no way to answer from its own recollection of a standard. The only way it obtains a control is by asking the graph for it, so what it tells you can be traced to a record.

13
tools it can call
10
steps per run, capped
Professional
tier required

Everything it is allowed to do

The complete list, read from the agent's own tool table at render time. It has no other way to act, and nothing here is aspirational.

search_frameworks

Search for frameworks by name or keyword. Use this to find exact framework names before calling other tools.

get_framework

Get details about a specific framework including its domains and control counts. Framework name must be exact.

get_framework_controls

List all controls in a framework, optionally filtered by domain.

get_control_detail

Get full details for a specific control including requirements and clauses.

get_control_cross_references

Get all controls that map to/from a given control across frameworks.

cross_framework_map

Get all control-to-control mappings between two frameworks.

gap_analysis

Find controls in framework1 that have NO mapping to framework2. Returns unmapped (gap) controls.

coverage_report

Get a coverage report showing how many controls in a framework map to each other framework.

search_controls

Full-text search across controls by keyword. Optionally filter by frameworks.

get_control_insights

Get real-world BPA subject insights relevant to a control.

create_remediation_task

Create a remediation task for a compliance gap. Saves to the user's portfolio.

save_finding

Save a control note/finding to the user's portfolio.

search_document

Search within an uploaded policy document for specific topics or keywords. Use this to find relevant policy sections when mapping against a framework.

What it is not

It does not touch your systems. It reads the graph, and where you have uploaded a document, that document. It cannot change a configuration, open a ticket in your tracker, or collect evidence out of a cloud account.

It is capped at 10 steps per run on purpose. An agent given an unbounded budget on a graph this size will keep finding one more thing to look at.

See the tier that includes itOr point your own agent at the graph