ScanMalware.com URL Scanner MCP server
2026년 11월 13일 커밋ScanMalware.com API를 사용하여 AI 클라이언트에서 URL 스캔, 악성코드 탐지 및 보안 분석을 수행합니다.
다른 이름 scanmalware/mcp-server, mcp-server
- 스타
- 1
- Tools
- —
- 다운로드, 30일
- —
- 최신 버전
- 0.2.2
- 마지막 커밋
- 2026년 11월 13일
- 라이선스
- Apache-2.0
ScanMalware.com URL Scanner MCP란?
ScanMalware.com API를 사용하여 AI 클라이언트에서 URL 스캔, 악성코드 탐지 및 보안 분석을 수행합니다.
프로젝트 README에서 생성된 요약
ScanMalware.com URL Scanner MCP server 정보
보안 분석가와 개발자는 ScanMalware.com API를 통해 AI 클라이언트에서 직접 심층 URL 스캔 및 악성코드 탐지를 수행할 수 있습니다. 이 Python 기반 MCP server는 스캔 요청, 완료 대기, 그리고 위험 점수, AI 분석, 스크린샷 리소스와 같은 상세 보안 인텔리전스 조회 기능을 제공합니다. 또한 다양한 가시성 모드를 지원하여 대상 URL의 민감도에 따라 공개, 미등록 또는 프라이빗 스캔 결과 중 하나를 선택할 수 있습니다.
서버는 streamable HTTP를 사용하여 로컬에서 실행하거나 Docker를 통해 실행할 수 있으며, SSE와 stateless HTTP transport를 모두 지원합니다. 고급 보안 요구 사항이 있는 경우, Nginx와 TLS 검사를 위한 mitmproxy가 포함된 특수 번들을 사용하여 DigitalOcean droplet에 배포할 수 있습니다. 사용자는 다운로드 및 JSON 응답에 대한 특정 제한을 구성할 수 있으며, SCANMALWARE_BEARER_TOKEN을 제공하여 프라이빗 스캔 데이터에 접근할 수 있습니다. 또한 JavaScript 핑거프린트 패턴, 난독화 및 악성코드 패밀리에 대한 특화된 검색 기능도 포함되어 있습니다.
활용 사례
- 1피싱 분류의심스러운 URL에 대한 스캔을 요청하고 위험 점수 및 주요 지표 요약을 확인합니다.
- 2브랜드 오용 모니터링기존 스캔 데이터에서 특정 브랜드 키워드를 검색하여 위험도가 높은 사기 사이트를 식별합니다.
- 3인증서 검사특정 스캔 ID에 대한 TLS 상세 정보 및 인증서 정보를 조회하고 요약합니다.
프로젝트의 README를 바탕으로 작성되었습니다. 모든 활용 사례는 해당 내용의 인용문과 대조하여 확인합니다.
ScanMalware.com URL Scanner MCP 설치 방법
형식 확인: 2026년 11월 10일Cursor에 추가Goose에 추가Claude Code에 ScanMalware.com URL Scanner MCP 설치
$ claude mcp add --transport http mcp-server https://mcp.scanmalware.com/mcpCursor에 ScanMalware.com URL Scanner MCP 설치
~/.cursor/mcp.json
{
"mcpServers": {
"mcp-server": {
"url": "https://mcp.scanmalware.com/mcp"
}
}
}VS Code에 ScanMalware.com URL Scanner MCP 설치
.vscode/mcp.json
{
"servers": {
"mcp-server": {
"type": "http",
"url": "https://mcp.scanmalware.com/mcp"
}
}
}VS Code CLI에 ScanMalware.com URL Scanner MCP 설치
$ code --add-mcp '{"name":"mcp-server","type":"http","url":"https://mcp.scanmalware.com/mcp"}'Codex에 ScanMalware.com URL Scanner MCP 설치
$ codex mcp add mcp-server --url https://mcp.scanmalware.com/mcpCodex config.toml에 ScanMalware.com URL Scanner MCP 설치
~/.codex/config.toml
[mcp_servers.mcp-server] url = "https://mcp.scanmalware.com/mcp"
Gemini CLI에 ScanMalware.com URL Scanner MCP 설치
$ gemini mcp add -s user --transport http mcp-server https://mcp.scanmalware.com/mcpGemini settings.json에 ScanMalware.com URL Scanner MCP 설치
~/.gemini/settings.json
{
"mcpServers": {
"mcp-server": {
"url": "https://mcp.scanmalware.com/mcp",
"type": "http"
}
}
}구성
DROPLET_NAMEMCP_HOSTMCP_PORTMCP_TRANSPORTSSH_KEYS
| 이름 | 설명 | 유형 |
|---|---|---|
| DROPLET_NAME | — | |
| MCP_HOST | — | |
| MCP_PORT | — | |
| MCP_TRANSPORT | — | |
| SSH_KEYS | — |
자동 검사
2026년 11월 14일에 실행됨전체 보고서 →아직 인증되지 않았습니다. 서버는 80점 이상의 점수, 심각한 결함 없음, 그리고 검사에 충분한 데이터를 갖추어야 합니다. 검사 방식 · 점수를 높이기 위해 수정할 항목
- 소스 저장소 연결 및 접근 가능scanmalware/mcp-server is public and reachablegithub.com/scanmalware/mcp-server ↗통과
- 패키지가 이 저장소를 소스로 명시검사되지 않음
- 빌드 출처(Sigstore 증명)검사되지 않음
- 오픈 소스 라이선스 (OSI)Apache-2.0spdx.org/licenses/Apache-2.0.html ↗통과
- 최근 1년 내 커밋 또는 릴리스Last commit or release 2026-11-12github.com/scanmalware/mcp-server ↗통과
- OpenSSF Scorecard검사되지 않음
- 패키지 이름이 인기 패키지와 유사하지 않음검사되지 않음
- 패키지 및 해당 종속성에 중요한 보안 권고 없음검사되지 않음
- 모델에 대한 숨겨진 지침 없음0 tool descriptions and the README read; no override phrases, credential paths or hidden characters통과
- 파괴적인 도구에 라벨 표시됨검사되지 않음
- 원격 엔드포인트가 MCP 핸드셰이크에 응답initialize answered with protocol 2025-06-18 in 97 ms통과
README
GitHub에서 보기scanmalware-mcp
Minimal Python MCP server that wraps the public ScanMalware.com API.
Operations
See the operations runbook for deployment, TLS, logging, security maintenance, and how to connect to the DigitalOcean droplet.
Run locally (Streamable HTTP)
python -m venv .venv
source .venv/bin/activate
pip install -U pip
pip install .
export MCP_TRANSPORT=streamable-http
export MCP_HOST=127.0.0.1
export MCP_PORT=8000
scanmalware-mcp
Run with Docker
docker build -t scanmalware-mcp .
docker run --rm -p 127.0.0.1:8000:8000 \
-e MCP_TRANSPORT=streamable-http \
-e MCP_HOST=0.0.0.0 \
-e MCP_PORT=8000 \
scanmalware-mcp
Optional: set MCP_AUTH_TOKEN to require Authorization: Bearer <MCP_AUTH_TOKEN> for HTTP transports.
Optional auth env vars (only needed for auth-gated endpoints):
SCANMALWARE_BEARER_TOKEN
Other env vars:
SCANMALWARE_BASE_URL(default:https://scanmalware.com)SCANMALWARE_ALLOW_HTTP(default:false)SCANMALWARE_TIMEOUT_S(default:30)SCANMALWARE_SLOW_QUERY_TIMEOUT_S(default:90; favicon and technology statistics, OCR text search, JS fingerprint similarity counts and behavioural signature search)SCANMALWARE_MAX_DOWNLOAD_BYTES(default:10485760)SCANMALWARE_MAX_JSON_BYTES(default:10485760; decoded JSON response limit)SCANMALWARE_MAX_ERROR_BYTES(default:16384; error body sample, with truncation marked)SCANMALWARE_ALLOW_PRIVATE_TARGETS(default:false)SCANMALWARE_CA_CERT(optional; path to a CA bundle for SSL bump)
MCP server security env vars:
MCP_AUTH_TOKEN(if set, HTTP transports requireAuthorization: Bearer <token>)MCP_RESOURCE_SERVER_URL/MCP_ISSUER_URL(optional; only used whenMCP_AUTH_TOKENis set)
Transport note: the HTTP transports run stateless (stateless_http=True), so responses carry
no mcp-session-id header. Clients must not require one. This keeps per-session state -
and therefore memory - flat. Both MCP protocol generations are served: initialize
(2024-11-05 to 2025-11-25) and server/discover (2026-07-28). Requests use POST;
GET on /mcp returns 405 because the server offers no standalone SSE stream.
SSE transport: set MCP_TRANSPORT=sse, and optionally MCP_SSE_MOUNT_PATH to serve
it under a path prefix.
Tool note: submit_scan does not call /api/v1/csrf-token; there is no CSRF token tool.
Tool note: some upstream endpoints are disabled and excluded from the tool list (e.g., get_improvements, find_screenshot_duplicates, get_ai_stats, search_js_fingerprinter2_code_hash, search_js_segments_by_tlsh).
Filters: every search tool can be called with its defaults.
search_js_fingerprint_patterns applies has_eval=true when no pattern filter
is given, because the API requires one; search_js_obfuscation and
search_js_malware_families fall back to the API's own defaults. Score filters
use the API's scales and are checked before the request: the AI risk score is
0-10, AI confidence 0-100, JS obfuscation scores 0-1 and the runtime JS risk
score 0-100. OCR search requires at least three characters after trimming.
Library lookups use detected identifiers from get_js_library_inventory;
search_js_fingerprint_by_library accepts Next.js as an alias for the
detected identifier nextjs.
Semantic search compares text with screenshot embeddings. Its default threshold
is -0.2, with a supported range of -1 to 1; useful cross-modal thresholds
are usually around -0.3 to 0.
wait_for_scan bounds the whole operation, including upstream requests, retry
backoff and sleeps. timeout_s must be greater than zero and at most 3600;
poll_interval_s must be between 0.1 and 60. Retryable failures use backoff and
honor Retry-After within that deadline. A timeout returns the last summary,
or status: "unknown" and an empty summary if no poll succeeded. Submissions
and other writes are not automatically retried.
Scan visibility
submit_scan requires scan_type; there is no default. A public scan puts the
target URL and scan results in ScanMalware's public feed, visible to other users
and search engines. Choose deliberately for client targets, confidential URLs,
and security engagements:
public: publishes the scan. Use for these targets only with explicit approval to publish.unlisted: excluded from public listings, but accessible to anyone with the direct link.private: results are restricted to the authenticated ScanMalware account. Requires a validSCANMALWARE_BEARER_TOKENconfigured on the MCP server;MCP_AUTH_TOKENonly controls access to the MCP server and does not provide ScanMalware authentication.
If visibility has not already been specified for these targets, ask the user
to choose unlisted or private before submitting. If submission fails, do not
retry with a less restrictive visibility. These modes control scan visibility;
requests still go to ScanMalware and may be recorded in server logs (see
the MCP privacy policy). ScanMalware documents the visibility
options in its privacy policy.
Example prompts
Phishing triage (submit → wait → summarize):
Submit an unlisted scan for https://example-login-update.com, wait for completion, and
return status, risk_score, and the top indicators. If high risk, include the
AI analysis and screenshot resource.
Brand abuse monitoring:
Search scans for "acme login" (limit 5). For each result, list scan_id,
status, risk_score, and URL. Highlight anything marked high risk.
TLS/certificate inspection:
For scan_id 1234...abcd, fetch TLS details and the certificate PEM download.
Summarize issuer, subject, validity dates, and SANs; flag mismatches.
Deploy to DigitalOcean (Debian + Docker + Nginx)
The deploy bundle lives in deploy/ and runs three containers:
mcp(this server, streamable HTTP on port 8000)nginx(frontend on ports 80/443; redirects HTTP to HTTPS and proxies/mcpto the MCP server)proxy(mitmproxy, checks ScanMalware destinations and logs upstream HTTP traffic)
Prereqs
doctlauthenticated (doctl auth init)- SSH key uploaded to DigitalOcean (used by
doctl compute droplet create) - DNS and Let's Encrypt certificates at the paths in HTTPS / TLS before starting Nginx
Create a small droplet in Germany (Frankfurt)
DROPLET_NAME=scanmalware-mcp-small
REGION=fra1
SIZE=s-1vcpu-2gb
IMAGE=debian-12-x64
SSH_KEYS=$(doctl compute ssh-key list --format ID --no-header | paste -sd, -)
doctl compute droplet create "$DROPLET_NAME" \
--region "$REGION" \
--size "$SIZE" \
--image "$IMAGE" \
--ssh-keys "$SSH_KEYS" \
--tag-name scanmalware-mcp \
--wait
Firewall (public HTTP/HTTPS + SSH)
doctl compute firewall create \
--name scanmalware-mcp-fw \
--inbound-rules "protocol:tcp,ports:22,address:0.0.0.0/0,address:::0/0" \
--inbound-rules "protocol:tcp,ports:80,address:0.0.0.0/0,address:::0/0" \
--inbound-rules "protocol:tcp,ports:443,address:0.0.0.0/0,address:::0/0" \
--outbound-rules "protocol:icmp,ports:0,address:0.0.0.0/0,address:::0/0" \
--outbound-rules "protocol:tcp,ports:0,address:0.0.0.0/0,address:::0/0" \
--outbound-rules "protocol:udp,ports:0,address:0.0.0.0/0,address:::0/0" \
--droplet-ids <droplet-id>
Install Docker + compose on the droplet
Install Docker Engine and the Compose plugin from the official Docker apt repository. Use docker compose; the old Python docker-compose client is incompatible with current Docker Engine.
See the operations runbook for pinned builds and maintenance checks.
Upload and run
# Package exactly one commit; a tar of the working tree also picks up untracked
# files (.tools/, caches, egg-info) and uncommitted edits.
git archive --format=tar.gz -o /tmp/scanmalware-mcp.tar.gz origin/main
scp -i /path/to/key /tmp/scanmalware-mcp.tar.gz root@<droplet-ip>:/tmp/
ssh -i /path/to/key root@<droplet-ip> \
"mkdir -p /opt/scanmalware-mcp && tar -xzf /tmp/scanmalware-mcp.tar.gz -C /opt/scanmalware-mcp"
On a fresh host, follow TLS inspection to build the proxy and generate its CA before starting MCP and Nginx. Install the egress rules and Docker startup hook and log rotation as part of host setup.
Verify
curl -I https://mcp.scanmalware.com/
curl -sS -o /dev/null -w '%{http_code}\n' https://mcp.scanmalware.com/mcp
/ should return 200 from Nginx. /mcp returns 405 on GET, which is expected: the server offers no standalone SSE stream, so MCP requests use POST.
Smoke test (MCP initialize + tools/list)
python - <<'PY'
import json
import httpx
URL = "https://mcp.scanmalware.com/mcp"
HEADERS = {
"accept": "application/json, text/event-stream",
"content-type": "application/json",
}
init_payload = {
"jsonrpc": "2.0",
"id": 1,
"method": "initialize",
"params": {
"protocolVersion": "2025-06-18",
"capabilities": {},
"clientInfo": {"name": "mcp-smoke-test", "version": "0.1.0"},
},
}
with httpx.Client(timeout=10) as client:
init_resp = client.post(URL, headers=HEADERS, json=init_payload)
init_resp.raise_for_status()
# Stateless deployments return no session header.
session_id = init_resp.headers.get("mcp-session-id")
def extract_sse_data(text: str) -> dict:
for line in text.splitlines():
if line.startswith("data: "):
return json.loads(line[len("data: "):])
raise ValueError("No SSE data line found")
init_message = extract_sse_data(init_resp.text)
protocol_version = init_message["result"]["protocolVersion"]
headers = {**HEADERS, "mcp-protocol-version": protocol_version}
if session_id:
headers["mcp-session-id"] = session_id
# Send initialized notification
client.post(
URL,
headers=headers,
json={"jsonrpc": "2.0", "method": "notifications/initialized"},
)
tools_resp = client.post(
URL,
headers=headers,
json={"jsonrpc": "2.0", "id": 2, "method": "tools/list"},
)
tools_resp.raise_for_status()
tools_message = extract_sse_data(tools_resp.text)
tool_names = [tool["name"] for tool in tools_message["result"]["tools"]]
print("protocol_version:", protocol_version)
print("tool_count:", len(tool_names))
print("tools:", ", ".join(tool_names))
PY
Redeploy / new deploys
Two common flows:
- In-place update (same droplet)
# Package exactly one commit; a tar of the working tree also picks up untracked
# files (.tools/, caches, egg-info) and uncommitted edits.
git archive --format=tar.gz -o /tmp/scanmalware-mcp.tar.gz origin/main
scp -i /path/to/key /tmp/scanmalware-mcp.tar.gz root@<droplet-ip>:/tmp/
ssh -i /path/to/key root@<droplet-ip> \
"bash /opt/scanmalware-mcp/deploy/redeploy.sh /tmp/scanmalware-mcp.tar.gz"
The redeploy script validates and builds in a staging directory while the current
stack runs, then switches files and waits for readiness. It preserves .env,
deploy/.env, the mitmproxy CA and logs. A failed build leaves the current stack
running; a failed startup restores the prior files and image tags. Other
untracked files are removed on a successful release. See the
runbook for persistent configuration and recovery.
Optional one-shot helper from the repo root (it packs the working tree, including uncommitted changes, rather than a commit):
./deploy/push-redeploy.sh root@<droplet-ip> /path/to/key
To update only the MCP server with a few seconds of downtime, see Updating only the MCP server.
- Rolling deploy (new droplet)
- Create a new droplet (steps above)
- Deploy the same bundle
- Switch DNS to the new IP
- Destroy the old droplet when ready
doctl compute droplet delete <old-droplet-id> --force
ScanMalware.com URL Scanner MCP 질문
ScanMalware.com URL Scanner MCP를 Claude Code에 추가하려면 어떻게 하나요?
터미널에서 claude mcp add --transport http mcp-server https://mcp.scanmalware.com/mcp를 실행합니다. 모든 프로젝트에서 사용할 수 있게 하려면 --scope user를 추가합니다.
ScanMalware.com URL Scanner MCP를 Cursor에 추가하려면 어떻게 하나요?
~/.cursor/mcp.json를 열고 mcpServers 아래에 서버를 추가합니다. 위 설치 섹션에 표시된 Cursor 구성을 사용합니다.
서버에서 프라이빗 스캔을 지원하나요?
네, 하지만 프라이빗 스캔을 위해서는 MCP server에 유효한 SCANMALWARE_BEARER_TOKEN이 설정되어 있어야 합니다.
Docker를 사용하여 이 서버를 실행할 수 있나요?
네, transport 및 host 환경 변수를 지정하여 Docker로 서버를 빌드하고 실행할 수 있습니다.