Skip to content

Back to the catalog

Strix
github.comrepo

Strix

Strix is an open-source framework of autonomous AI pentesting agents that dynamically run target code, discover vulnerabilities across the OWASP Top 10 (SQLi, SSRF, XXE, IDOR, XSS, auth bypass, business-logic flaws) and validate them with working proof-of-concept exploits rather than static-analysis heuristics. It ships a CLI with a full offensive toolkit (Caido-based HTTP interception, browser exploitation, Python exploit sandbox, recon/OSINT, SAST/DAST), supports OpenAPI/Postman-driven black-box testing, integrates into GitHub Actions/CI pipelines to block vulnerable PRs, and can run against local or hosted LLMs via LiteLLM, with an optional cloud offering for autofix PRs and compliance reporting.

TL;DR Open-source multi-agent AI pentesting CLI that autonomously exploits and validates web/app vulnerabilities with real PoCs, plus CI/CD gating.
Open original ↗
Document

What it is

Strix's setup and operational surface, beyond the capability list the summary covers. The tool runs as a Docker-sandboxed agent driven by your own LLM API key, with three delivery tiers: open-source (local, self-hosted), Strix Cloud (app.strix.ai), and Enterprise. The summary tells you what it detects; this tells you what it takes to run it and where results land.

Preconditions

  • Docker running locally (first run auto-pulls a sandbox image).
  • An LLM API key from any supported provider — OpenAI, Anthropic, Google, or others via OpenRouter. This is bring-your-own-key; the open-source tier has no vendor inference cost but you pay your provider per run.

What to do

Install and run a first assessment:

# Install Strix
curl -sSL https://strix.ai/install | bash

# Configure your AI provider
export STRIX_LLM="openrouter/z-ai/glm-5.3"
export LLM_API_KEY="your-api-key"

# Run your first security assessment
strix --target ./app-directory

Results are written to strix_runs/<run-name> as the run proceeds. Review them in a local dashboard:

# Open the most recent run
strix view

# ...or open a specific run by name
strix view my-run-name

# Expose the viewer on all IPv4 interfaces at a fixed port
strix view --host

To wire it into an existing coding agent (Claude Code, Cursor, Codex, any SKILL.md-compatible), which installs nine skills covering pentests, fixes, and CI scanning:

npx skills add usestrix/strix

References: AGENTS.md (quick reference), docs.strix.ai/llms.txt (CLI), docs.app.strix.ai (API).

Caveats

  • curl … | bash piping an install script and strix view --host (binds the results viewer to all IPv4 interfaces) are both worth scrutinising before use on a shared or exposed host.
  • Auto-fix (ready-to-merge PR patches), continuous pentesting, one-click autofix, and DevSecOps integrations (GitLab, Bitbucket, Slack, Jira, Linear) are described as Cloud/Enterprise features — the page does not claim they are in the open-source CLI. Compliance reports (SOC 2, ISO 27001, PCI DSS), SSO, and BYOK/tailored agents are Enterprise-only.
  • The tool executes and exploits targets dynamically inside the sandbox; point it only at systems you are authorised to test.

Technologies named here

Techniques named here

Related in Offensive Security