AdStrikeAdStrike by capture0x
A terminal Active Directory attack framework with an AI operator mode, MCP tool server, Smart Analyst, and session-aware evidence workflow for authorized assessments.
Engagement workspace
AdStrike stores target, domain, credential, Kerberos, finding, command, BloodHound, and runtime state in one workspace. Its modules and AI paths reuse that context instead of forcing repeated operator input.
Capabilities
These capabilities summarize how AdStrike connects discovery, identity analysis, AI assistance, execution paths, and reporting into one assessment workflow.
Keep target, domain, username, password or NT hash, Kerberos ccache, findings, command history, BloodHound data, and runtime artifacts in one reusable session.
Module map
AdStrike groups its menu into nine kill-chain phases plus utilities, covering no-credential surface checks, identity abuse, lateral movement, credential access, persistence, cloud and hybrid paths, reporting, Smart Analyst, and the AI agent.
DNS, OSINT, nmap, masscan, IPv6, surface discovery
NTLM capture, relay paths, CVE checks, AMSI/EDR/UAC workflows
LDAP, SMB, GPO, DNS, BloodHound, NetExec/NXC, user hunting
Kerberos, ADCS, ACL abuse, RBCD, Shadow Credentials, RODC
PSExec, WMIExec, SMBExec, DCOM, WinRM, MSSQL
LSASS, SAM, NTDS, DPAPI, DCSync, shadow copies
Tickets, AdminSDHolder, GPO, DNSAdmins, trusts, local persistence
Azure AD, Entra hybrid, ADFS, Golden SAML, M365, Teams
Exploit chains, C2 integration, loot parsing, AD playbooks, AI agent
MCP integration
AdStrike can be connected to MCP-capable AI clients so the client supplies the model while AdStrike provides structured actions for the assessment.
Store the domain controller, domain, identity, and authentication method once, then reuse that context across the assessment without repeating sensitive values in every request.
claude mcp add adstrike -- /path/to/AdStrike/venv/bin/python3 /path/to/AdStrike/mcp_server.pyThe MCP path exposes the same assessment capabilities used by the standalone AI operator, including discovery, identity analysis, Kerberos workflows, execution paths, planning, and reporting.
Screenshots
Quick start
$ git clone https://github.com/capture0x/AdStrike.git
$ cd AdStrike
$ chmod +x install.sh run.sh
$ bash install.sh
$ source venv/bin/activate
$ bash run.sh
$ python3 main.py --check
Analyst and reports
The agent can run full-auto or plan-only with Ollama or Claude. Smart Analyst then reads collected artifacts and reports turn findings, command history, and attack-chain reasoning into reviewable output.
Run with Ollama local models or Claude API models, choose full-auto execution or plan-only mode, and select loud, normal, or stealth OPSEC behavior.
Smart Analyst reads BloodHound JSON, LDAP enumeration files, session findings, agent logs, Kerberos artifacts, hashes, and collected evidence before building a prioritized attack plan.
Agent runs write round-by-round Markdown logs, recorded tool commands, sanitized results, mission summaries, captured-hash placeholders, findings, and attack-chain plans.
FAQ
AdStrike is built for authorized Active Directory assessments where the operator needs discovery, identity analysis, Kerberos workflows, execution paths, evidence handling, and reporting in one terminal workflow.
No. The manual console and modules can be used directly. AI-assisted operation is available for users who want planning, full-auto workflows, or MCP-driven operation.
It lets an MCP-capable AI client call structured AdStrike actions while AdStrike keeps the engagement context and assessment output organized.
No. It supports lab and professional assessment workflows, but it must only be used on systems you own or have explicit written authorization to test.
AdStrike can preserve command history, findings, collected artifacts, live agent notes, and reports in HTML, Markdown, and JSON formats.
The workflows focus on Active Directory environments, including Windows domain services, Kerberos, LDAP, SMB, ADCS, BloodHound data, WinRM, MSSQL, cloud and hybrid identity paths.
AdStrike is intended only for lawful administration, research, and security testing on systems you own or have explicit written authorization to assess.