MX CyberOps Agent Platform

Defensive, enterprise-grade AI-assisted cybersecurity operations platform for regulated environments.

Current version: 0.6.1

Purpose

MX CyberOps Agent Platform is a read-only-first SecOps agent framework that normalizes alerts, enriches identity/device/network context, builds incident timelines, produces typed evidence packs, and routes any impactful action through policy and approval gates.

Safety Boundary

This repository is designed for defensive operations only:

• Read-only mode by default.
• No arbitrary shell execution.
• No autonomous destructive action.
• No exploitation tooling.
• No unbounded remediation.
• Every tool request is policy-evaluated.
• Every agent decision is auditable.
• Jira/JSM output is draft-only unless a future approval-gated workflow is added.
• Write-like connector operations are denied by read-only connector contracts.

Current Capabilities

• SOC triage agent
• Identity security agent
• Defensive hunt/query generation agent
• Microsoft Graph connector for Entra sign-in/audit logs and Intune device compliance
• Okta System Log connector
• Tanium asset/client-health connector scaffold
• Jira/JSM draft-ticket connector
• Unified enrichment pipeline
• Typed enrichment evidence export
• Policy engine
• Prompt-injection firewall
• Audit logger
• FastAPI service
• CLI runner
• Mock mode by default, live connectors opt-in by env vars

Quick Start

python -m venv .venv
source .venv/bin/activate
pip install -e .[dev]
pytest -q

Run sample enrichment in mock mode:

mxcyber enrich \
  --upn exec@example.com \
  --hostname SURFACE-FLOOR-001 \
  --incident-id SEC-2026-0001

Run API:

uvicorn apps.api.main:app --reload

Docker lab:

docker compose up --build

API Endpoints

GET  /health
GET  /connectors
GET  /connectors/health
POST /connectors/{name}/query
POST /enrichment/run
GET  /hunts
POST /hunt/generate
POST /identity/assess
POST /identity/aggregate   # backward-compatible; delegates to enrichment pipeline
POST /policy/evaluate

Typed Evidence Output

The enrichment CLI writes typed artifacts under:

runtime/evidence/<incident_id>/<run_id>/

Artifact manifest:

identity-events.json
device-posture.json
tanium-health.json
jira-draft.json
connector-health.json
identity-assessment.json
timeline.json
evidence-refs.json
warnings.json
manifest.json
executive-summary.md

Live Connector Environment

All live connectors are disabled unless explicitly enabled.

MX_GRAPH_ENABLED=true
MX_GRAPH_TENANT_ID=<tenant-id>
MX_GRAPH_CLIENT_ID=<app-client-id>
MX_GRAPH_CLIENT_SECRET=<client-secret>

MX_OKTA_ENABLED=true
MX_OKTA_ORG_URL=https://example.okta.com
MX_OKTA_API_TOKEN=<read-only-token>

MX_TANIUM_ENABLED=true
MX_TANIUM_BASE_URL=https://tanium.example.com
MX_TANIUM_API_TOKEN=<read-only-token>

MX_JIRA_ENABLED=true
MX_JIRA_BASE_URL=https://example.atlassian.net
MX_JIRA_EMAIL=analyst@example.com
MX_JIRA_API_TOKEN=<token>
MX_JIRA_PROJECT_KEY=SEC

References

• NIST Cybersecurity Framework 2.0: https://www.nist.gov/cyberframework
• NIST AI Risk Management Framework: https://www.nist.gov/itl/ai-risk-management-framework
• CISA Secure by Design: https://www.cisa.gov/securebydesign
• OWASP Top 10 for LLM Applications: https://owasp.org/www-project-top-10-for-large-language-model-applications/
• MITRE ATT&CK: https://attack.mitre.org/