Usage
Configure the endpoint, credential references, and TLS policy before invoking a client or server. See Configuration.
Python client
from ciso_assistant_api.auth import get_client
client = get_client()
try:
response = client.api_incidents_list(status="open", limit=25)
records = response.data
finally:
client.close()
Every generated method returns a Response wrapper with decoded .data, the
underlying HTTP response, status, and headers. List operations follow bounded
DRF pagination automatically.
Library callers may pass a token directly to get_client as an in-memory value.
Durable deployments must use secret-reference settings instead.
MCP server
Run the condensed surface locally:
ciso-assistant-mcp --transport stdio
Each domain router accepts an action and a JSON-encoded params_json string.
For example, invoke ciso_assistant_incidents with:
{
"action": "api_incidents_list",
"params_json": "{\"status\":\"open\",\"limit\":25}"
}
Retrieve exact source objects before correlating them:
{
"action": "api_risk_scenarios_retrieve",
"params_json": "{\"id\":\"<source_uuid>\"}"
}
The generated operation inventory in the README is authoritative for available
actions. Keep MCP_TOOL_MODE=condensed for delegated agents; enable verbose
per-operation tools only for an approved diagnostic need.
Knowledge-graph ingestion
ciso_ingest materializes supported source records with provenance. Start with a
bounded sample and state the target graph before invoking it:
{
"kind": "risk_scenarios",
"params_json": "{\"limit\":100}"
}
Supported kinds include risk scenarios, risk assessments, applied controls, vulnerabilities, compliance assessments, incidents, and evidence metadata. The MCP ingestion boundary never persists raw evidence attachment bytes.