Open Autonomous SecOps · Vendor-Independent, transparent, auditable, and accessible to all

Democratizing Autonomous Defense

Traditional SOCs operate at human speed against AI-enabled adversaries attacking at machine speed. Proprietary AI SOCs cost fortunes and hide inside black boxes. ZeroSOC is the open ecosystem standardizing autonomous security operations for all.

Democratizing Defense Executor-Neutral OCSF-Aligned Glass-Box Auditable Community Governed

An Open Working Draft for Practitioners

No single team or vendor has all the answers for autonomous security operations. The ZeroSOC project is an open working draft (v0.1) uniting specifications, agent reasoning skills, MCP tool servers, and runtime harnesses under transparent community governance. We invite practitioners and security leaders to critique, test, and co-develop with us.

The Operational Challenges

Why SecOps Needs an Open Foundation

Modern security operations face structural challenges that proprietary black boxes fail to resolve.

The Missing Foundation

Lack of Operational Standards

SecOps lacks universal operational semantics: different teams attribute conflicting meanings to basic concepts. Organizations constantly reinvent the wheel authoring siloed procedures from scratch. Deploying AI agents on unstandardized processes produces non-deterministic decisions and inconsistent execution.

Speed Asymmetry

Machine-Speed Adversaries

Adversaries increasingly automate reconnaissance, craft dynamic payloads, and compress dwell times to minutes. Manual triage operating at human speed cannot keep pace with automated attack chains.

Cost & Lock-In

Commercial Pricing Barriers

Proprietary "AI SOC" solutions attach luxury enterprise pricing floors, opaque data-ingest surcharges, and vendor lock-in, leaving most organizations unable to deploy modern SecOps automation.

Trust Deficit

Opaque "Black-Box" Automation

Closed AI platforms obscure their reasoning logic and prompt chains. Security teams cannot safely delegate containment actions without verifiable audit trails, inspectable evidence, and human supervision.

Architecture & Workstreams

How the ZeroSOC Ecosystem Fits Together

The ZeroSOC initiative is organized into five focused open workstreams that connect specifications, agent reasoning, native security tools, runtime execution, and continuous empirical validation into a cohesive ecosystem.

End-to-End Ecosystem Architecture
01 · Specs
Framework
Taxonomy & Playbooks
02 · Reasoning
Agent Skills
Cognitive Routines
03 · Tooling
MCP Servers
Direct Native APIs
04 · Runtime
Platform
Execution & UI Console
05 · Eval
Cyber Range
Adversary Benchmark
Normative Standard

The ZeroSOC Framework Architecture

The Framework provides the foundational body of knowledge structured into seven core modules and an auditable, NIST-aligned, 4-phase operational lifecycle.

ZeroSOC Framework Infographic showing 7 core modules
Figure 1. ZeroSOC Framework Architecture — 7 Core Modules & Operational Interconnections
Principle 01

Aspirational North Star, Pragmatic Reality

Autonomy is a continuum, not a binary switch. Real-world SecOps balances deterministic scripts for repetitive volume, AI agents for judgment under uncertainty, and human analysts for safety gates, ambiguous escalations, and accountability.

Principle 02 · Karpathy's Law

Executor Neutrality & Human Readability

"You can outsource thinking, but you cannot outsource understanding." Processes are independent of executor type—a human, script, or AI agent follows the exact same normative steps in clean Markdown.

Principle 03

Closed-Loop Continuous Feedback

Triage false-positives and post-incident root cause analyses feed directly back into detection engineering baselines and playbook updates, systematically eliminating alert rot.

Detection & Response Lifecycle

The 4 Phases & 5 Measurement Gates

The operational pipeline standardizes 4 phases governed by 5 measurement gates (G1–G5) with explicit I/O contracts. Click any phase below to inspect its inputs, outputs, and actor roles.

Phase 2 · Detection & Analysis

Triage & Investigation

G2 & G3 — Triage Decision & Verdict: OCSF 2005
Objective & Roles

Fast alert aggregation into Cases, domain-specific triage, and concurrent A/B hypothesis testing.

Primary Actors: Incident Investigator, Threat Hunter, AI Orchestrator
Inputs (Consumes)
  • Alerts (OCSF 2004) aggregated into Cases (OCSF 2005)
  • Enrichment sources (CMDB, Identity, Threat Intel)
  • Standardized Playbooks (04-Playbooks)
Outputs (Produces)
  • G2: Closed Case OR Promotion to Investigation
  • G3: Confirmed Incident + Category (IC-##)
  • Triage Note & Investigation Note
Reference Runtime & UI Console

The ZeroSOC Platform

An open-source execution engine and analyst console designed for glass-box reasoning visibility, local perimeter execution, and strict human-in-the-loop control.

ZeroSOC Analyst Console: Glass-Box Investigation Trace

Glass-Box Investigation Tracing

Every query, LLM reasoning step, and concurrent A/B hypothesis evaluation (Malicious vs. Benign) is recorded in a transparent, inspectable timeline for full auditability and regulatory compliance.

ZeroSOC Analyst Console: Operator Autonomy Controls

Operator Autonomy & Guardrails

Operators configure confidence thresholds and mandatory Human-in-the-Loop (HITL) approval gates per incident category, ensuring containment actions stay strictly bounded within organizational policy.

Public Kickoff · September 15, 2026

Join the ZeroSOC Working Group

On September 15, 2026, we are launching a formal public Call to Action for SecOps practitioners, detection engineers, incident responders, and security leaders to expand the community working group and organize the first collaborative development sprints.

Playbooks & OCSF

Author and peer-review domain triage and A/B hypothesis investigation playbooks mapped to real-world ATT&CK techniques and incident types.

MCP Tool Servers

Build and enhance Model Context Protocol servers connecting SecOps agents safely to native SIEM, EDR, and Cloud APIs.

Range & Benchmarks

Create adversary emulation scenarios, synthetic telemetry sets, and evaluation tests to benchmark autonomous SecOps reasoning.

Join GitHub Discussions & Pre-Register Star on GitHub
Primary Hub Working Group Slack (Launching Sept 15)
RFCs & Specs GitHub Discussions & Issues
Practitioner FAQ

Frequently Asked Questions

Addressing the hard questions and real skepticism regarding autonomous security operations.

No. We do not believe fully autonomous ("lights-out") SOCs are achievable or desirable today. ZeroSOC represents an aspirational north star to guide collective practitioner research and standardization.

In reality, mature SecOps is a hybrid continuum: deterministic automation for routine volume, AI agents for bounded hypothesis testing under uncertainty, and human analysts for safety gates, ambiguous escalations, and accountability.

This is one of the most critical open questions facing modern cybersecurity. Historically, practitioners developed intuition, tradecraft, and deep system understanding by spending years grinding through entry-level alert triage.

While the ZeroSOC North Star aims to replace repetitive manual analysis with autonomous operations, seasoned human SMEs remain essential for human-on-the-loop oversight, safety gates, detection engineering, threat modeling, and novel multi-stage incidents. If automated systems handle routine triage, the traditional apprenticeship model must evolve.

We believe future analysts will learn by auditing transparent reasoning traces, authoring open playbooks, and training in empirical cyber ranges—supervising and validating AI agents rather than performing repetitive manual lookups. We actively invite SecOps leaders, educators, and practitioners to join this community discussion to shape the future of cybersecurity career paths together.

Structure precedes autonomy. You cannot automate what you haven't standardized. Giving an AI agent access to production tools without formal taxonomy, explicit schema contracts, and normative playbooks leads to unpredictable hallucinations and inconsistent remediation — risking accidental business disruption on one side, or incomplete threat containment on the other.

The ZeroSOC Framework provides the foundational operational grammar: Karpathy's Law ("you can outsource thinking, but you cannot outsource understanding"), executor-neutral playbooks, and verifiable G1–G5 measurement gates.

Security through obscurity is not security. Relying on hidden, informal operational procedures leaves defenders with unvetted blind spots and untested assumptions.

Just as MITRE ATT&CK standardized adversary behaviors and Sigma standardized detection rules, open and peer-reviewed playbooks ensure that investigation hypotheses, evidence checks, and triage paths are intrinsically robust, auditable, and continuously validated against real attack simulations.

ZeroSOC follows an open-core licensing posture designed to maximize community adoption while defending the open runtime:

  • Apache-2.0: zerosoc-framework, zerosoc-skills, zerosoc-mcp, and zerosoc-cyber-range are permissively licensed so specifications, agent skills, tool servers, and benchmarks spread freely without friction.
  • AGPLv3: zerosoc-platform (the reference execution engine) uses copyleft to ensure that downstream runtime improvements remain open to the entire community.

Unconstrained autonomy is dangerous. That is why ZeroSOC mandates safe-by-default tool separation (read tools strictly isolated from write/response tools), explicit autonomy sliders, and mandatory Human-in-the-Loop (HITL) approval gates for any non-reversible action.

Transparent reasoning logs and case-attributed audit trails ensure that teams maintain complete visibility and control over automated actions.

Contributions follow an open, issue-first workflow with transparent licensing and contributor provenance:

  1. Open an Issue / RFC First: Discuss proposals on GitHub to align on scope.
  2. Author & Format: Write Markdown playbooks, MCP tools, or cyber range scenarios using designated templates.
  3. Verify via Tabletop / Range: Validate actions empirically before promotion.
  4. Peer Review & Merge: Collaborate on GitHub pull requests under clear open-source licensing (Apache-2.0 / AGPLv3).