AdStrike has emerged as a game‑changing framework for authorized Active Directory assessments, blending modular design with cutting‑edge AI capabilities. Unlike monolithic tools that force testers to juggle disparate utilities, AdStrike centralizes session state—credentials, Kerberos tickets, discovered objects, and executed commands—within a shared JSON context. This eliminates repetitive data entry and reduces human error during long engagements. By storing everything in one place, operators can seamlessly pivot between enumeration, exploitation, and reporting while maintaining a clear audit trail. The framework’s terminal‑based interface remains lightweight yet powerful, making it suitable for both lab environments and enterprise‑scale assessments where precision and repeatability are paramount.

The true innovation lies in AdStrike’s MCP (Model Context Protocol) server, which exposes all 53 tools—52 attack modules plus a session‑setter—without requiring any API key, local model download, or Ollama installation. Instead, it leverages the LLM subscription already present in popular MCP hosts such as Claude Code, Cursor, or Claude Desktop. This architecture shifts the computational burden to the host’s AI, allowing the framework to focus purely on tool orchestration and result parsing. For teams already invested in Claude‑based workflows, the integration is frictionless: simply point the host to the AdStrike MCP server and let the model drive the engagement loop, invoking tools like nmap_scan, enumerate_ldap, or kerberoast in a logical sequence based on real‑time output.

AdStrike’s kill‑chain coverage spans nine distinct phases, each represented by dedicated menu ranges that guide testers from initial reconnaissance through advanced operations such as C2 integration and exploit chaining. The framework includes specialized modules for Kerberos attacks, ADCS certificate abuse, Shadow Credentials, RBCD chains, and even cloud/hybrid vectors like Azure AD and M365. This breadth ensures that regardless of the target’s hardening posture—whether NTLM is disabled, LDAP signing is enforced, or MFA is in place—there is a corresponding primitive within AdStrike to probe, exploit, or bypass the control. The modular nature also encourages community contributions, continuously expanding the primitive library beyond the current 400+ tradecraft entries.

Kerberos‑aware workflows are a cornerstone of AdStrike’s effectiveness in modern, hardened AD environments. When NTLM is unreliable or blocked, the framework can auto‑generate a tailored krb5.conf, inject DC FQDN mappings into /etc/hosts, and obtain a TGT using Impacket’s getTGT.py. It then sets the necessary KRB5CCNAME and KRB5_CONFIG environment variables, enabling subsequent modules—NetExec, BloodHound, Evil‑WinRM—to operate seamlessly in Kerberos‑only mode. Additionally, the TGT_AUTO_RENEW flag keeps tickets fresh without manual intervention, while the Smart Analyst parses Kerberos‑specific output (e.g., service ticket names, encryption types) to prioritize next steps such as AS‑REP roasting or constrained delegation abuse.

The optional AdStrike Agent brings AI‑assisted planning directly into the framework, offering two modes: Full Auto, where the agent autonomously selects and executes tools based on evolving evidence, and Plan Only, which generates a prioritized attack roadmap without taking action. This flexibility satisfies both operators who prefer hands‑on control and those who wish to offload repetitive decision‑making to an LLM. Complementing the agent, the built‑in Smart Analyst module performs real‑time parsing of command output, ranks discovered assets by relevance, and suggests the next most promising module—essentially acting as a force multiplier that reduces cognitive overload and speeds up the OODA loop during time‑boxed engagements.

Reporting is another area where AdStrike excels, offering automated generation of HTML, Markdown, and JSON reports that capture every executed command, raw output, and contextual findings. Crucially, the framework respects operational security by default: the ADSTRIKE_SHOW_SECRETS flag masks passwords, hashes, and loot in logs and reports unless explicitly overridden. All artifacts are stored under an output/ directory, neatly segregated into subfolders for enumeration, BloodHound data, agent logs, and temporary Kerberos material. This structure simplifies post‑engagement review, redaction, and handover to stakeholders, ensuring that sensitive information never inadvertently leaks while still delivering a comprehensive narrative of the assessment’s findings and impact.

Getting started with AdStrike is deliberately straightforward to avoid common pitfalls. The quick‑start sequence—git clone, chmod +x install.sh run.sh, bash install.sh, source venv/bin/activate, bash run.sh—creates a isolated virtual environment and installs core dependencies without requiring sudo, thereby preventing permission‑related file ownership issues. Should a user accidentally run the installer with elevated privileges, a simple chown command restores proper ownership. The framework also provides a Tool Checker (module 58) and a repair script (scripts/repair_tools.sh) to validate and remediate missing external utilities such as Impacket, NetExec, Certipy, or BloodHound, ensuring that the operator’s environment stays ready for action.

Beyond the core installer, AdStrike demands a modest set of external tools that mirror the standard red‑team arsenal: impacket‑scripts, nxc/netexec, bloodhound‑python, certipy‑ad, evil‑winrm, kerbrute, responder, ldap‑utils, hashcat, john, nmap/masscan, and krb5‑user. Most of these are pulled in automatically, but the repair script can reinstall or upgrade them when version mismatches occur. Network prerequisites focus on reachability to typical AD ports (88, 389, 443, 445, 636, 5985), and the framework can optionally perform a quick nmap port check during session startup to confirm connectivity before diving into deeper enumeration or attack phases.

From a market perspective, AdStrike reflects a broader shift toward AI‑augmented offensive security platforms that prioritize automation, contextual awareness, and integration with existing developer toolchains. As enterprises harden their identity infrastructures with Azure AD, Conditional Access, and MFA, traditional password‑spraying or pass‑the‑hash techniques lose efficacy, driving demand for frameworks that can intelligently chain Kerberos, certificate, and cloud‑based attack vectors. AdStrike’s MCP‑first approach aligns with the rise of “AI‑as‑a‑service” in security tooling, where the model resides in the user’s preferred IDE or assistant, and the security framework acts as a pluggable toolset—reducing duplication, lowering the barrier to entry, and fostering collaborative workflows between red teams, blue teams, and developers.

For security professionals looking to adopt AdStrike, the following actionable steps will maximize value while maintaining compliance and safety. First, always obtain explicit written authorization before targeting any AD environment; the framework’s disclaimer and licensing (GPLv3) emphasize authorized use only. Second, begin with the recommended first‑run flow: use the Session Manager to define target credentials, run the Tool Checker to validate dependencies, perform baseline AD Enumeration, let the Smart Analyst prioritize next steps, and finally generate a report to document findings. Third, leverage the MCP server if you already use Claude Code, Cursor, or Claude Desktop—this eliminates the need for separate API keys and lets your existing LLM subscription drive the engagement. Fourth, monitor the ADSTRIKE_OPSEC and ADSTRIKE_SHOW_SECRETS flags to ensure logs and reports remain sanitized for sharing with stakeholders or inclusion in internal knowledge bases. Finally, contribute back to the community by sharing custom modules, reporting false positives, or suggesting new primitives; the framework’s open‑source nature thrives on collaborative improvement, ensuring it stays ahead of evolving AD defenses.