Stankye

AssemblyLine 4 MCP Server

Community Stankye
Updated

AssemblyLine 4 MCP Server

MCP server for the AssemblyLine 4 malware analysis platform.

Setup

npm install
npm run build

Copy .env.example to .env and fill in your credentials:

AL4_URL=https://your-al4-instance:443
AL4_USERNAME=your_username
AL4_APIKEY=keyname:your_api_key_password

API keys are created in AssemblyLine under Account → Security → Manage API Keys.

Running

AL4_URL=... AL4_USERNAME=... AL4_APIKEY=... node dist/src/index.js

Claude Code integration

Add to your ~/.claude/claude_desktop_config.json (or project .claude/settings.json):

{
  "mcpServers": {
    "assemblyline4": {
      "command": "node",
      "args": ["/absolute/path/to/vibe-assemblylinev4-mcp/dist/src/index.js"],
      "env": {
        "AL4_URL": "https://your-al4-instance:443",
        "AL4_USERNAME": "your_username",
        "AL4_APIKEY": "keyname:password"
      }
    }
  }
}

Self-signed / dev certificates

Set AL4_TLS_VERIFY=false in the env block to skip TLS verification.

Available tools

Tool Description
al4_whoami Current authenticated user
al4_submit_file Sync file submission (path on disk)
al4_submit_url Sync URL submission
al4_submit_sha256 Sync submission by hash (file must exist in AL4 store)
al4_ingest_file Async file ingestion with optional notification queue
al4_ingest_url Async URL ingestion
al4_ingest_sha256 Async hash ingestion
al4_submission_is_complete Poll whether a submission has finished
al4_submission_get Submission metadata
al4_submission_full Complete results tree
al4_submission_summary Summarised results
al4_ingest_get_messages Drain a notification queue
al4_search_submissions Lucene search over submissions
al4_search_alerts Lucene search over alerts
al4_search_files Lucene search over the file store
al4_search_results Lucene search over service results
al4_alert_get Fetch a single alert by ID
al4_file_info File metadata by SHA256
al4_file_results All service results for a file
al4_file_score Highest score for a file

Testing & CI

Two layers of tests live under test/:

  • test/integration.ts — drives AL4Client and the MCP server end-to-endagainst an in-process mock of the AL4 REST API (test/mock-al4.ts).Fast, hermetic, runs on every push/PR via .github/workflows/ci.yml.

    npm test
    
  • test/e2e.ts — drives AL4Client against a real Assemblyline 4appliance. The .github/workflows/e2e.yml workflow spins up the upstreamassemblyline-docker-composeappliance (core stack only — Elasticsearch, Redis, MinIO, AL core/UI/nginx;service images are skipped to fit the runner's disk quota), bootstraps theadmin user, mints an API key, and runs the suite. Because submissions arenot blocked on services, they complete with an empty result tree — enoughto validate every API the MCP server exposes.

    The e2e workflow runs:

    • nightly (05:27 UTC),
    • on workflow_dispatch,
    • on PRs labelled run-e2e.

    To reproduce locally:

    bash scripts/ci/start-al4.sh
    APIKEY=$(AL4_URL=https://localhost AL4_ADMIN_USER=admin AL4_ADMIN_PASSWORD=admin \
             bash scripts/ci/create-apikey.sh)
    AL4_URL=https://localhost AL4_USERNAME=admin AL4_APIKEY=$APIKEY \
      AL4_TLS_VERIFY=false node dist/test/e2e.js
    

Running CI locally

Two equivalent paths, depending on what you want to validate:

  1. The fast workflow, exactly as GitHub Actions runs it — viaact:

    # one-time: install act (https://nektosact.com/installation/)
    npm run ci:local            # runs .github/workflows/ci.yml in Docker
    

    act reads .actrc, which pins the catthehacker ubuntu:act-latestimage (has Node, git, curl, etc.). This is the cleanest way to catchregressions in the workflow YAML itself.

  2. The e2e workflowact is not recommended here because the jobspins up its own docker-compose stack, and docker-in-docker via actmakes networking and disk usage painful. Use the wrapper scriptinstead, which runs the same steps directly on your host:

    npm run e2e:local           # build → start appliance → mint key → run e2e → teardown
    KEEP_RUNNING=1 npm run e2e:local   # leave appliance up afterwards for poking
    SKIP_BUILD=1   npm run e2e:local   # skip npm ci + tsc, reuse dist/
    

    Or break it apart:

    npm run e2e:up              # just bring up the appliance
    APIKEY=$(npm run --silent e2e:apikey)
    AL4_URL=https://localhost AL4_USERNAME=admin AL4_APIKEY=$APIKEY \
      AL4_TLS_VERIFY=false npm run test:e2e
    

    Requirements: Linux host (or WSL2 / macOS), docker + compose plugin,node 20+, ~15 GB free disk. Windows users: run the scripts from insideWSL — the appliance won't come up under Git Bash because of howdocker-compose paths and openssl interact on native Windows.

MCP Server · Populars

MCP Server · New

    sjkim1127

    Reversecore_MCP

    A security-first MCP server empowering AI agents to orchestrate Ghidra, Radare2, and YARA for automated reverse engineering.

    Community sjkim1127
    sebringj

    Autonomo MCP

    Tired of 'it works' lies? Autonomo MCP makes your AI prove it—on real hardware, right in your editor.

    Community sebringj
    softerist

    Heuristic MCP Server

    Enhanced MCP server for semantic code search with call-graph proximity, recency ranking, and find-similar-code. Built for AI coding assistants.

    Community softerist
    arm

    Arm MCP Server

    Arm's MCP server

    Community arm
    bobmatnyc

    MCP Vector Search

    CLI-first semantic code search with MCP integration. Modern, fast, and intelligent code search powered by ChromaDB and AST parsing.

    Community bobmatnyc