MCP tools

emfirge_scan

Build the read-only cloud graph that your agent can fork and evaluate across modeled consequences.

Parameters

role_arn: string · required; region: string · required

MCP ask / arguments
{
  "role_arn": "arn:aws:iam::123456789012:role/EmfirgeReadOnly",
  "region": "us-east-1"
}

Returns

The response is JSON shaped for the operation. The fields and semantics are described below.

JSON response
{
  "analysis_id": "7f3c9a2e-4b1d-4c8a-9f2e-1a2b3c4d5e6f",
  "region_analyzed": "us-east-1",
  "overall_risk_score": 3,
  "overall_risk_level": "CRITICAL",
  "security_score": 18,
  "availability_score": 96,
  "disaster_recovery_score": 90,
  "cost_score": 97,
  "simulation_baseline": { "critical_count": 15 },
  "critical_risks": [ /* … full finding objects … */ ],
  "moderate_risks": [ /* … */ ],
  "low_risks": [ /* … */ ],
  "toxic_combinations": [ /* … */ ],
  "total_resources_scanned": 51
}

Validation, behavior, and limits

Returns an overall_risk_score (0–100, higher is safer), matching overall_risk_level, per-category security_score, availability_score, disaster_recovery_score, and cost_score, plus analysis_id, region_analyzed, simulation_baseline, grouped critical_risks, moderate_risks, and low_risks arrays, toxic_combinations, and total_resources_scanned.

Resource IDs are tokenized to NAME_### by default. The score is deterministic and read-only; there is no separate counts object because tallies live under simulation_baseline. Hold on to analysis_id for get_findings, attack_paths, verify_fix, simulate_breach, and check_compliance.

The result includes the remaining daily quota when available. A real scan takes 30–60 seconds and streams to stay under proxy timeouts. The default limit is five scans per account per local calendar day (override with DAILY_SCAN_LIMIT); 429 means the limit was reached. Other tools use a completed scan and do not consume this budget.

Based on the Emfirge MCP and engine source.View source
Documentation