Your AI SOC,
Open Source.

A coordinated team of 13 specialized AI agents in your SOC. Built on the Claude Agent SDK and MCP.

bash — vigil
$ git clone --recurse-submodules https://github.com/Vigil-SOC/vigil.git
$ cd vigil && ./start_web.sh
✅ Agents initializing...
✅ 30+ MCP integrations discovered
✅ 7,200 detection rules loaded
Open http://localhost:6988 — your AI SOC is running.
13
Specialized Agents
30+
MCP Integrations
7,200+
Detection Rules
$0
License Cost (Apache 2.0)
< 3 min
Time to Running SOC
⚠️ A Word About the "AI SOC" Market

The year is 2025. A venture-backed startup wraps a Claude API call in some middleware, slaps "Agentic AI SOC" on the homepage, and charges you $40,000–$200,000 a year to start. The agent reasoning? A black box. The integrations? Proprietary. Your data? Uploaded to their cloud. When you ask why an agent made a call? "Trust us." We're looking at you AI SOC venders. You know who you are and what you did. The silver bullet cycle isn't working. It's time to build the capability, not buy the box. That's Vigil.

Open Source AI SOC

What Is Vigil?

Vigil is an open-source AI security operations platform. It puts a coordinated team of 13 specialized AI agents to work as your SOC, each with deep access to your security stack, each purpose-built for a specific part of the investigation lifecycle.

Built on the Claude Agent SDK and the Model Context Protocol (MCP), Vigil is designed around a simple principle: AI agents should do the work, not just answer questions.

Say "Run incident response on finding XYZ" and agents execute in sequence: triage scores the alert, the investigator traces root cause, the responder submits containment actions, and the reporter generates audit-ready documentation. No hand-offs required.

Apache 2.0 licensed. Every agent's reasoning is inspectable. Every workflow is a text file you can read, modify, and share.

View on GitHub Quick Start →
Vigil web dashboard — AI SOC interface
The Problem

Why Open Source?

The cybersecurity market has a well-documented structural problem. Closed-source AI SOC platforms deepen it.

🔍

Transparency

When agent reasoning is inspectable, detection quality becomes measurable. You know why an agent made a call, not just that it did.

🔗

Open Integrations

Built on MCP—an open standard. When integrations are built on open standards, the ecosystem grows the platform. Every new MCP server is a free Vigil integration.

🤝

Collective Knowledge

When skills are text files, the community's collective operational knowledge compounds. Share a playbook and everyone's SOC gets smarter.

"Security operations shouldn't be a black box you buy. It should be a capability you build, together."

Architecture

How Vigil Works

Three interlocking layers that give every agent real-time access to your entire stack.

⚙️ Skills — Complete Workflows in a Single Command

A skill is a multi-agent workflow that chains specialized agents into a complete, end-to-end playbook. Each skill maps to a real operational workflow and produces real outputs. Creating your own skill is writing a SKILL.md file. If your team has a workflow, Vigil can run it.

🚨Incident Response
Triage Investigate Respond Report

End-to-end from alert to audit-ready documentation.

🔬Full Investigation
Deep Dive Correlate Timeline

Deep-dive reconstruction with correlation across all connected data sources.

🎯Threat Hunt
Hypothesis Hunt ATT&CK Map

Proactive hypothesis-driven search using MITRE ATT&CK as the framework.

⚖️Forensic Analysis
Collect Chain-of-Custody Report

Evidence collection and analysis.

🔌 MCP — Your Entire Stack, Connected

Vigil uses the Model Context Protocol to give every agent real-time access to your existing tools. 30+ integrations out of the box. If a tool has an MCP server, Vigil can connect to it.

Splunk
CrowdStrike
VirusTotal
Shodan
Hybrid Analysis
Jira
Slack
DeepTempo
+ 22 more
Write your own → Any MCP server plugs right in

🏗️ Architecture

Skills Layer
SKILL.md Files — Multi-Agent Workflows
Composable playbooks. Define agent sequence, phase-level tool access, and natural-language instructions. Skills can invoke other skills as sub-workflows.
Agent Layer
Claude Agent SDK — 13 Specialized Agents
Defined roles, tool access, and reasoning modes. The SDK handles coordination, state management, and tool invocation. 23 backend tools + 100+ extended tools via MCP.
Integration Layer
MCP — 30+ Tool Integrations
Any MCP-speaking tool or data source plugs in without custom code. Handles authentication, rate limiting, and response normalization. Open standard — ecosystem grows the platform.
The Team

13 Agents. Each Built for a Job.

Specialists with defined roles, tuned reasoning modes, and access to 23 backend tools plus 100+ extended tools via MCP. You set the automation thresholds. Vigil keeps humans in the loop where it counts—and gets out of the way everywhere else.

T Triage Agent ⚡ Fast mode
I Investigator 🧠 Deep mode
M MITRE Analyst
C Correlator
R Responder 🔒 Threshold routing
📄 Reporter
D Detection Engineer
🔎 Enrichment
👤 Identity Analyst
🌐 Network Forensics
🦠 Malware Analyst
📋 Compliance Mapping
📂 Case Management

🔒 The Confidence Threshold: The Responder auto-approves containment actions at 0.90+ confidence. Below that, it routes to a human reviewer. You control the automation boundary. No surprises. No runaway automation.

🛡️

7,200+ Detection Rules

Spanning Sigma, Splunk, Elastic, and KQL formats. AI-assisted coverage analysis, gap identification, and template generation. Every new community rule improves every deployment.

💬

Chat-Driven Case Management

Build and update cases through natural language. Tell the system a finding is part of the lateral movement kill chain—it handles the MITRE tagging, timeline updates, and case linkage automatically.

🔁

Headless Autonomous Mode

Run Vigil as a daemon for 24/7 monitoring without touching the UI. Use ./start_daemon.sh for always-on coverage, or the full React + FastAPI frontend for interactive investigations.

🏠

Local-First Architecture

Your data never leaves your environment. No cloud dependency for core functionality. MCP connections are under your control. State persists locally in SQLite (or Postgres for team deployments).

60-Second Demo

See It Work in Under a Minute

After starting Vigil, open the web UI and try the full incident response workflow live—using sample data included in the repo.

1

Paste a finding

Use a real alert ID from your SIEM, or one of the included sample findings from the repo.

2

Run the skill

Type: Run incident response on finding f-20260215-abc123

3

Watch agents execute

Triage → Investigate → Respond → Report. Every step visible, every decision inspectable.

4

Review the output

Complete incident report with MITRE ATT&CK mapping, timeline, and recommended actions. Audit-ready.

The Honest Comparison

Vigil vs. Commercial AI SOCs

Every serious AI SOC platform on the market is closed source. Here's the full picture.
(Including Dropzone AI, Conifers CognitiveSOC, Radiant Security, Prophet Security, Exaforce, Torq HyperSOC. None are open source. None use MCP.)

⚔️ Vigil (Open Source) Commercial AI SOCs
License Apache 2.0 — free forever Proprietary, $36K–$200K+/yr
Source Code Fully inspectable on GitHub Closed / opaque
Agent Logic Transparent, modifiable Python Black box or patented
Integrations MCP (open standard), 30+ Proprietary APIs, 50–100+
Extensibility Write a SKILL.md file Vendor roadmap or pro services
Data Residency 100% local, your machine Cloud APIs, data leaves your env
Time to Try git clone, < 3 minutes Sales call, 30-day POC, procurement
Community Open contributions welcome Feature requests into a backlog
Detection Rules 7,200+ included, community-maintained Proprietary or third-party subscription
LLM Backend Claude (default), extensible architecture Vendor-locked to one provider
Community

Built to Be Extended

Vigil is a platform, not a product. Here's what you can build—no vendor permission required.

📝

Write a New Skill

Skills are markdown files defining multi-agent workflows. Phishing triage, cloud incident response, insider threat—encode any playbook and share it with the community. No code required.

🔌

Build an MCP Integration

If your security tool has an API, wrap it in an MCP server and Vigil connects to it. SIEM, threat intel, firewalls, identity providers—all welcome.

🛡️

Add Detection Rules

Contribute rules in Sigma, Splunk SPL, Elastic KQL, or any format. Every new rule improves the Detection Engineer agent for everyone.

🤖

Improve an Agent

Each agent's behavior is readable Python. Better MITRE mapping, tighter forensic chain-of-custody formatting, sharper triage scoring—open a PR.

🗺️

Roadmap

Expanded MCP Integrations

Sentinel, Elastic, Palo Alto Cortex, Google Chronicle, AWS Security Hub.

Community Skills Catalog

An npm for SOC workflows—searchable by use case, alert type, and tool stack.

Additional LLM Backends

OpenAI, local models via Ollama, and other providers.

Federated Deployment

Run agents across multiple environments—data stays local to each.

Detection-as-Code Pipeline

Version-controlled detection rules with automated testing and deployment.

Auth is WIP

Default is a Dev mode that bypasses it for now.

Quick Start

Up and Running in 3 Minutes

Three commands. No sales call. No procurement. No 30-day POC.

Requirements

Python 3.11+
Node.js 18+
Docker Desktop
Claude API key (optional for initial testing)

Installation

$ git clone --recurse-submodules \
https://github.com/Vigil-SOC/vigil.git
$ cd vigil
$ ./start_web.sh
 
# First run: 2–3 minutes
# Subsequent starts: <30 seconds
Frontend: http://localhost:6988
API Docs: http://localhost:6987/docs

DeepTempo + Vigil

Vigil is sponsored by DeepTempo. DeepTempo's LogLM connects to Vigil via MCP as a high-fidelity behavioral detection layer— giving you AI-native detections on top of an open AI SOC. Vigil works with Splunk alone. Adding DeepTempo means AI-native behavioral detection on top of it.

Clone Vigil on GitHub
VIGIL

Join the Community

Vigil is early. The architecture is solid, the agents are working, and the skills system is production-tested. What we need now is the community, people who run SOCs, build integrations, write detections, and know what's broken in their security operations today.

The agents are transparent. The skills are text files. The integrations are open standard. Everything is designed for you to make it yours.