
Self-hosted AI security center with no API key required: AiSOC, up 446 stars, Agent inference can be replayed.
Open-source, self-hosted AI Security Operations Center (AiSOC): A rule library of 7023 rules, LLM agent triage, full decision traceability, no API key required, and built-in CISA KEV intelligence.
Related tags
The security team's biggest concern about using large models for security operations wasn't their capabilities, but the "black box" nature of the models: if a model says an alert is a false positive, why? The AiSOC project, with 6263 stars, addresses this issue with a "replayable decision log" – every prompt and tool call made by the AI agent is recorded verbatim, allowing for retrospective review. This MIT-licensed, self-hosted AI security operations center (SOC) gained 446 stars today, making it trending on GitHub.
From Monitoring to Remediation Advice: Full Workflow
AiSOC operates as a complete pipeline: after receiving security tool telemetry data, it first normalizes the data and then runs against a rule library – with 7023 rules, 2538 are executable. When an alert is triggered, it's grouped into an event, which is then investigated by the AI agent. Every inference process of the agent is fully recorded, ultimately resulting in remediation advice – but any action must be approved by a human before being implemented.
Even more impressive is the readily available threat intelligence: the built-in intelligence module directly loads real CISA KEV vulnerability directories, which can be used immediately after booting up, without requiring any API keys. The new version of threat intelligence also re-scans historical data that has been collected.
Product Details: Transparent and Honest
AiSOC puts a lot of effort into transparently presenting the system status: it includes a local model for automated triage, displaying the original text of the model's determination, confidence level, and rationale on the interface. The SOC operations panel displays an empty state when no tools are connected, rather than showing a placeholder image. The repository also includes a 27-page technical guide PDF, architecture documentation, and an audit document titled "What Actually Works" – proactively telling users which features are actually available.
Getting Started and Integration
Deployment requires only two commands: git clone https://github.com/beenuar/AiSOC && cd AiSOC && make up. Windows and bare-metal environments offer one-click installation scripts (install.ps1 / install.sh), which also install all the necessary dependencies. The current version is v21.0.0.
For teams that want to integrate AI into their security operations but are hesitant to relax audit requirements, AiSOC includes an MCP server that can directly connect to tools like Claude, Cursor, and Continue – making it easier for engineers to incorporate AI into their daily conversations.
Repository URL: https://github.com/beenuar/AiSOC
More News
Want to see more quality content?
Browse all news