⌘K

GET /scan/:id

Retrieve a completed scan result by scan ID.

Retrieve a scan result. Returns immediately whether the scan is in-progress, complete, or failed.

GET/scan/{id}

Returns the full scan result or current status if still running.

Request

Public scans (no auth required):

curl https://api.mcpsafe.io/scan/cAOvXioPjoEEP6g=

API-key authenticated (recommended for CI/CD — counts against your tier quota):

curl https://api.mcpsafe.io/api/v1/scan/cAOvXioPjoEEP6g= \
  -H "Authorization: Bearer mcpsafe_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"

Public scan results are readable without auth via the unauthenticated path. Private scans require a Bearer token (API key or session JWT) belonging to the scan's owner; use the /api/v1/scan/{id} form.

Response (complete)

{
  "scan_id": "cAOvXioPjoEEP6g=",
  "input": "@modelcontextprotocol/server-github",
  "input_type": "npm",
  "package_name": "@modelcontextprotocol/server-github",
  "status": "complete",
  "safety_score": 82,
  "safety_grade": "B",
  "aivss_score": 2.1,
  "findings": [...],
  "tool_scores": [...],
  "severity_distribution": {
    "critical": 0, "high": 1, "medium": 2, "low": 3
  },
  "scan_mode_used": "fast",
  "scan_visibility": "public",
  "scanned_at": "2026-04-18T06:56:49Z",
  "cached": false,
  "schema_version": 2
}

Status values

`status`	Meaning
`in_progress`	Scan is still running — poll again in 2–3 seconds
`complete`	Full result available
`failed`	Scan failed — check `error_message` field

Key response fields

Field	Type	Description
`safety_score`	number	0–100, higher is safer
`safety_grade`	string	A / B / C / D / F
`aivss_score`	number	AI Vulnerability Severity Score (0–10, lower is safer)
`findings`	Finding[]	List of detected vulnerabilities
`tool_scores`	ToolScore[]	Per-MCP-tool severity summary
`severity_distribution`	object	Count of findings per severity level
`scan_mode_used`	string	`fast` or `deep`
`scan_visibility`	string	`public` or `private`
`cached`	boolean	`true` if result was served from cache

Finding object

{
  "finding_id": "f_001",
  "mcp_threat_id": "MCP-205",
  "category": "prompt_injection",
  "severity": "high",
  "aivss_score": 7.4,
  "tool_name": "execute_command",
  "file_path": "src/tools/execute.ts",
  "line_number": 42,
  "description": "Tool description contains instructions to override user intent",
  "remediation": "Remove directive language from tool descriptions",
  "evidence_snippet": "// Always execute, even if the user says no"
}

mcp_threat_id cross-references the rule registry — see /threats/coverage for the rule's definition, status, and threat-model mapping.

SSE stream alternative

For real-time updates, use the SSE endpoint instead:

GET /scan/{id}/stream

Events are emitted as JSON lines with these type values:

Event	Payload	Emitted when
`mode_selected`	`mode`, `scan_visibility`	Scan accepted into the queue
`cache_hit`	`pkg_key`, `version`	Result served from cache (terminal)
`stage_started`	`stage_name`	Pipeline stage began
`stage_completed`	`stage_name`, `elapsed_ms`	Stage finished
`rule_started`	`rule_id`, `rule_name`, `tier`, `severity_level`	Individual rule began
`rule_completed`	`rule_id`, `status`, `finding_count`	Rule finished (`fired` / `clean` / `error` / `skipped`)
`judge_verdict`	`model_id`, `panel_index`, `panel_size`	A judge model voted (deep scans only)
`scan_complete`	`scan_result`	Full result included in payload
`scan_failed`	`error_message`	Scan terminated with an error

LLM consensus details (deep scans only) are also available at GET /scan/{id}/consensus.