Archsec-Emman/Agentic-Blue-Team

GitHub: Archsec-Emman/Agentic-Blue-Team

Stars: 1 | Forks: 0

403-Killchain Banner

# Agentic Blue Team **An open‑source, agent‑centric platform that fuses AI intelligence with a built‑in Security Incident Response Platform to automate alert triage, enrichment, and response — all within your own infrastructure.** ## 📖 Overview Agentic Blue Team is built to transform how security operations centers (SOCs) handle alerts. Instead of a static dashboard, the platform orchestrates modular AI agents (powered by local LLMs) to continuously ingest alerts, enrich them with threat intelligence, and produce structured security records. A fully integrated SIRP—built on the Nocoly worksheet platform—manages cases, alerts, artifacts, and playbooks, allowing analysts to trigger automated remediation or threat‑hunting workflows with a single click. The entire system is designed for local deployment: your data, models, and operations never leave your environment, ensuring complete confidentiality and compliance. ## ✨ Core Features | Feature | Description | |---------|-------------| | **AI‑Driven Intelligence** | Pre‑built agent templates (LangGraph, Dify) that use local LLMs to analyze alerts, enrich data, and determine outcomes. | | **Built‑in SIRP** | Full incident response platform on Nocoly with rapid customization of UIs, data models, reports, and workflows. | | **Powerful Automation Workflow** | Webhook‑based ingestion from Splunk / Kibana → Redis Streams → agent analysis → SIRP → playbooks. | | **Highly Extensible** | Rich library of Python modules and plugins for integrating with any security device or API. | | **Local Deployment & Data Control** | Fully self‑hosted; everything stays inside your environment. | | **Streaming + Batch Processing** | Real‑time alert analysis plus event‑driven automation for user‑triggered tasks. | ## 🧠 Architecture Agentic Blue Team follows a **five‑stage pipeline** that maps directly to the modern SOC workflow: SIEM (Splunk/ELK) → Webhook → Redis Stream → AI Agents → SIRP → Playbooks ### 1. Alert Ingestion Splunk, Kibana/ELK, or any SIEM forwards security alerts via a **webhook** to the platform’s built‑in receiver. ### 2. Stream Processing Alerts are pushed into dedicated **Redis Streams** (one per alert type), which act as persistent, replayable message queues. ### 3. AI Agent Analysis Modular AI agents consume alerts from the streams and perform deep analysis using **local LLMs**, threat intelligence enrichment, and rule‑based decision making. ### 4. SIRP Integration Agent outputs are formatted into standardized security records and sent to the built‑in **Security Incident Response Platform**, where they create or update cases, alerts, and artifacts. ### 5. Automated Response Analysts trigger **playbooks** directly from the SIRP interface to execute further actions — such as threat hunting, remediation, or notifications — closing the loop. ## 🗃️ Core Data Model The SIRP is built on a **clean, ontology‑driven data model** that separates detection, investigation, and response concerns. | Entity | Purpose | |--------|---------| | **Case** | The primary investigation object; aggregates alerts, artifacts, enrichments, and tickets. | | **Alert** | Middle‑layer object that retains detection‑source information (OCSF style) and links to artifacts. | | **Artifact** | The atomic unit: IP, domain, hash, username, etc. – the natural input for any query or enrichment. | | **Enrichment** | Structured, attachable information (threat intel, CMDB data, logs) that can be independently maintained. | | **Ticket** | Synchronization with external ticketing systems, linked to a Case. | | **Playbook** | Automated logic (often LangGraph orchestration) that runs against Cases, Alerts, or Artifacts. | | **Knowledge** | Shared, high‑currency knowledge base that serves both human analysts and AI agents. | ## 🚀 Installation ### Prerequisites - Python 3.10+ - Redis - Docker (optional, for containerized deployment) ### Quick Start # Clone the repository git clone https://github.com/Archsec-Emman/Agentic-Blue-Team cd Agentic-Blue-Team # Install dependencies pip install -r requirements.txt # Run database migrations python manage.py migrate # Start the development server python manage.py runserver The platform will be available at `http://localhost:8000`. ### Docker Deployment (Recommended for Production) # Build and start all services docker-compose up -d # Services: # - Web UI (Nocoly SIRP) # - Redis Streams # - Uvicorn (ASGI server) # - Ollama (optional, for local LLMs) ## ⚙️ Configuration ### Connecting a SIEM (Splunk / ELK) 1. Configure your SIEM to send alerts via webhook to `http://:8000/api/webhook/siem` 2. Set the appropriate authentication token in `settings.py`. 3. Alerts will automatically be routed to the correct Redis stream based on their type. ### Using Local LLMs Agentic Blue Team supports any local LLM through the **Ollama** plugin. To integrate: 1. Install Ollama and pull a model (e.g., `ollama pull llama3.2`). 2. In the platform, go to **Plugins → Ollama** and enter the model name. 3. Agents will automatically use the configured LLM for analysis. ### Customizing the SIRP The built‑in SIRP (Nocoly) allows full customization of: - **UIs** – Drag‑and‑drop dashboards and forms - **Data models** – Add custom fields to Cases, Alerts, Artifacts, etc. - **Reports** – Generate compliance or summary reports - **Workflows** – Define approval chains or automated transitions Refer to the `ARCHITECTURE.md` file for a detailed explanation of the data model and extension points. ## 🔌 Plugin Ecosystem Agentic Blue Team includes a growing library of **plug‑and‑play modules** under the `PLUGINS/` directory. | Plugin | Purpose | |--------|---------| | **AlienVaultOTX** | Pull threat intelligence from OTX | | **ClaudeCode** | Use Anthropic’s Claude for agent reasoning | | **ELK** | Native integration with Elastic Stack | | **Embeddings** | Vector embeddings for alert similarity search | | **Forwarder** | Relay alerts to external systems | | **Huggingface** | Access Hugging Face models | | **LLM** | Generic LLM adapter (OpenAI, Gemini, etc.) | | **MCP** | Model Context Protocol support | | **Qdrant** | Vector database for artifact correlation | ## 📂 Project Structure Agentic-Blue-Team/ ├── ABT/ # Django project configuration │ ├── settings.py # Main settings (plugins, LLMs, webhooks) │ ├── urls.py # URL routing │ └── asgi.py / wsgi.py ├── AGENTS/ # Modular AI agents │ ├── agent_cmdb.py # CMDB enrichment agent │ ├── agent_report.py # Report generation agent │ ├── agent_siem.py # SIEM alert processing │ └── agent_threat_intelligence.py ├── Core/ # Core data models (Case, Alert, Artifact, etc.) │ ├── models.py # SIRP data model │ ├── serializers.py │ └── views.py # REST API endpoints ├── PLAYBOOKS/ # Automated response playbooks │ ├── ALERT/ # Alert‑specific playbooks │ ├── ARTIFACT/ # Artifact‑specific playbooks │ └── CASE/ # Case‑level remediation workflows ├── PLUGINS/ # Extensible plugin library (see list above) ├── Lib/ # Core libraries (API, base modules, etc.) ├── DATA/ # Agent‑specific data handlers ├── Docker/ # Docker Compose services (DB, Redis, Ollama, etc.) ├── ARCHITECTURE.md # Detailed architecture reference ├── manage.py # Django management script ├── pyproject.toml # Python project metadata └── README.md # This file ## 📄 License This project is licensed under the **MIT License** – see the [LICENSE](LICENSE) file for details.