asmhunter-mcp

Local stdio MCP server for ASMHunter — drive your bug-bounty recon, monitoring, triage and reporting from an AI agent (Claude Code, Claude Desktop, Cursor, …).

What it is honestly good at: automating the ~80% grind — discovering fresh scope, spinning up targets, running recon + automated scans, reading back the attack surface and findings, organizing a hunt, and drafting reports. It puts you first on newly-added scope and never lets you miss a change. It does not find the high-value manual bugs for you (IDOR, business logic, auth bypass) — that's still your job; this tool buys you the time to do it.

Tools (34)

Intel — search_programs, get_program, recommend_programs, get_watchlist, get_scope_changes

Scanner (the ASMHunter edge) — list_targets, create_target, list_workflows, trigger_scan (pro+, confirm gate), get_scan_status, get_findings, list_assets

Hunt sessions — create_session, list_sessions, get_session, update_session, delete_session

Entries — add_lead, add_finding, add_tested, add_note, bulk_add_entries, list_entries, update_entry, delete_entry, promote_entry

Reports — draft_report, list_reports, update_report, delete_report

Stats & attribution — get_my_stats, get_value_receipt, list_wins, add_win

Requirements

• Python 3.12+
• An ASMHunter pro+ account (Personal Access Tokens and the scanner tools are pro-gated)
• A Personal Access Token — mint one at asmhunter.app → Settings → API Tokens (asm_live_…)

Install

uv tool install asmhunter-mcp
# or:  pip install asmhunter-mcp

Connect from Claude Code

Add via the CLI:

claude mcp add asmhunter \
  --env ASMHUNTER_TOKEN=asm_live_YOUR_TOKEN_HERE \
  -- uvx asmhunter-mcp

Or add to your MCP config (~/.claude/claude_desktop_config.json or Claude Code MCP settings):

{
  "mcpServers": {
    "asmhunter": {
      "command": "uvx",
      "args": ["asmhunter-mcp"],
      "env": {
        "ASMHUNTER_TOKEN": "asm_live_YOUR_TOKEN_HERE"
      }
    }
  }
}

ASMHUNTER_API_URL defaults to https://asmhunter.app — only set it to override (e.g. a self-hosted or staging instance).

Environment variables

Drive a hunt — example prompt

Once connected, in Claude Code just ask for the loop in plain language:

"Show me programs with recent scope changes. Pick one with genuinely fresh scope, add its new domain as a target, list the available scan workflows and run a recon + subdomain-enum scan (confirm it). When it finishes, summarize the newly-discovered assets and any findings by severity, flag which ones are worth manual testing, and draft a report skeleton for the most promising one."

Claude will chain get_scope_changes → create_target → list_workflows → trigger_scan(confirm=true) → get_scan_status → list_assets / get_findings → add_lead / draft_report. You then do the creative manual testing on the surface it prioritized. Ongoing, keep targets under continuous monitoring and let the alerts (Telegram/email) tell you when something changed.

Run directly

ASMHUNTER_TOKEN=asm_live_... asmhunter-mcp

Architecture

• client.py — httpx-only async client, no fastmcp dependency (unit-testable standalone; rejects path-traversal in interpolated ids)
• server.py — FastMCP tool wrappers (thin, no business logic; tiers/quota/RLS enforced server-side)
• __main__.py — entrypoint (mcp.run())

All tier enforcement, quota checks, and RLS live in the orchestrator — the MCP layer is a pure passthrough except for trigger_scan's local confirm gate.