Argus user guide

Argus is an open-source application security agent that reviews code the way an analyst does: it runs real scanners, reads their findings, weighs them against your organization’s context, and tells you in plain language what actually matters. Then you argue back — Argus answers.

It runs as one long-lived daemon per organization, reached where your team already works: a terminal chat, GitHub pull requests, and MCP from your own AI tools.

Argus runs in two shapes. Configure a model and it reasons on its own behalf, as above. Configure none and it is a toolbox: the same scanners, the same organizational knowledge, the same memory and the same methodologies, served to the AI tool you already pay for — which supplies the reasoning, on your machine, as you. Nothing about Argus requires a model budget to start.

These pages are task-oriented — how to install Argus, configure it, connect it, and host it.

• Deployment shapes — toolbox and colleague, and which one you are running
• Getting started — install, bootstrap, first chat
• Configuration — the full argus.yaml reference
• LLM providers — what Argus requires from a model, and how to verify yours
• GitHub channel — automatic pull-request reviews and talking to Argus on threads
• MCP channel — connect your own AI tools to Argus’s scanners, knowledge and skills
• Skills — using, writing, and overriding skills
• Kubernetes deployment — hosting Argus on a cluster