AutoMCP / CoAnalyst

An agentic LLM analyst for OT/IT alert triage, built on LangGraph and the Model Context Protocol. CoAnalyst reads security alerts out of an Elastic SIEM, correlates them against MITRE ATT&CK and threat intel, and can drive human-gated response and counter-reconnaissance against the source of an attack.

The implementation lives in KisteGraph/KisteAgent, see that directory's README for architecture, setup, and the security posture.

Publications

• WFCS 2026 (Work-in-Progress): CoAnalyst: An Agentic, LLM-based Cybersecurity Analyst for Industrial Control Systems, Çil, Rahmani, Sikora. Introduces the system.
• SHIELD-AI 2026 (ECML PKDD Workshop): When the Analyst Scans for the Attacker: Subverting Agentic LLM Security Operations in OT Networks, Rahmani, Çil, Sikora. Red-teams CoAnalyst's own agentic layer: three vulnerabilities in target selection, alert retrieval, and summarization, and one matched mitigation per vulnerability, evaluated on the live pipeline. The mitigations from that paper (M1-M3) are implemented in mitigations.py and wired into the deployed agent.

Repository layout

KisteGraph/KisteAgent/   The agent: LangGraph state machine, MCP tool server,
                          security mitigations, tests.

Everything else that used to live here (three standalone Assistants/ MCP servers and an earlier CLI prototype, security_assistant_v1/) predates the LangGraph rewrite and is superseded by KisteGraph/KisteAgent. That history is preserved at tag legacy-pre-cleanup-2026-07-09 rather than deleted outright.