Vigil queries Amazon Security Lake where the data sits: read-only Athena SQL over the OCSF sources Security Lake keeps in the Glue catalogue, one region per call. Regions, sources and OCSF versions are discovered from the Glue catalogue shared to the query access when the server starts, and the discovery results form part of the tool’s description. Sources are found by Security Lake naming (amazon_security_lake_glue_db_*, amazon_security_lake_table_*); a resource link renamed away from that naming is not found.

Tool

Tool Purpose
security_lake_query Run Athena (Trino) SQL naming sources by short name against one AWS region, over an ISO 8601 time range (start inclusive, end exclusive). A query_id argument instead collects the results of an earlier query in that region.

Vigil configuration

Configure in Settings → Integrations → Amazon Security Lake:

Field Notes
Role ARN Optional. A role to assume for querying, from the default AWS credentials of the host the server runs on.
External ID Optional. Sent when assuming the role above, if its trust policy requires one.
Athena Workgroup Default primary. The workgroup each query runs in, by the same name in every region, with a query result output location set.
Regions Returning Rows Comma-separated region codes whose result rows Vigil receives. Empty means none.

There is no access-key field: credentials come from the default AWS chain of the process the server runs in — an instance role, ECS task role, IRSA or Pod Identity, or a mounted shared config named by AWS_CONFIG_FILE / AWS_SHARED_CREDENTIALS_FILE with AWS_PROFILE. Under the MCP runner each server has its own HOME, so ~/.aws is not read. The backend runs the Settings connection test on its own chain, so the backend and the MCP runner must share a credential source: on Helm the same service account, on Compose the same instance role or the same mounted config.

Behaviour and limits

  • Allowed regions gate what Vigil receives, not what runs. A region not listed in Regions Returning Rows still runs each query and answers with its query id, row count and bytes scanned — and no rows.
  • Athena writes every query’s results to the workgroup’s output location in each region. That location, its lifecycle and the workgroup’s data-scan limit are the operator’s; the scan limit is the only cap on Athena spend.
  • Result rows pass Vigil’s credential redaction, which masks known credential field names (such as password, secretAccessKey, sessionToken) and token shapes only. A secret under any other name — for example an SSM parameter value or EC2 user data inside CloudTrail requestParameters or responseElements — can reach a run.