Find your next open-source project

Search by name, purpose, or keyword — or paste a GitHub repository URL

usestrix/strix

Security & Privacy

Autonomous AI agents that run dynamic penetration tests against code, web apps and APIs, validating findings with proofs-of-concept.

76,410
8,676
Python
Apache-2.0
1 year old

What it does

Strix runs autonomous AI pentesting agents against a local codebase, a GitHub repository, or a live web app or API. Rather than only flagging patterns like a static scanner, the agents execute code in a sandbox, probe for OWASP Top 10 and other vulnerability classes (IDOR, injection, SSRF, XSS, JWT and session flaws, API misconfiguration), and try to confirm each finding with a working proof-of-concept. Results are written to disk per run and can be inspected in a local dashboard.

It targets developers and security teams who want faster dynamic testing than manual engagements and fewer false positives than static analysis. Multi-agent orchestration lets specialist recon, exploitation and post-exploitation agents collaborate on the same target or across several targets.

Quick start

Prerequisites are a running Docker installation and an LLM API key from a supported provider. Install the CLI, point it at a model, and run a first assessment:

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

export STRIX_LLM="vercel_ai_gateway/zai/glm-5.3"
export LLM_API_KEY="your-api-key"

strix --target ./app-directory

The first run pulls the sandbox Docker image automatically, and output lands in strix_runs/<run-name>. Open the latest run in the browser viewer with strix view, or a named run with strix view my-run-name. Use strix --update to upgrade the standalone binary, or the matching pipx/uv/pip upgrade command if you installed that way.

Other entry points from the CLI reference: strix --target https://github.com/org/repo for a repository review, strix --target https://your-app.com for black-box testing, passing an OpenAPI/Swagger file or Postman collection alongside the live base URL for API coverage, --instruction for authenticated grey-box testing, and -n/--non-interactive for headless runs that exit non-zero when vulnerabilities are found. A GitHub Actions workflow can run quick scans on pull requests; it needs full history in checkout so changed-file scoping works GitHub Actions setup.

Configuration is saved to ~/.strix/cli-config.json. Besides metered API keys, Strix can sign in with a ChatGPT Plus/Pro subscription (strix auth login chatgpt), run against local models via LLM_API_BASE, or delegate to the managed cloud (strix cloud ..., no Docker or key required). Its own MCP servers can be connected through ~/.strix/mcp-servers.json. Coding agents such as Claude Code, Cursor and Codex can install Strix skills via npx skills add usestrix/strix.

Constraints and safety

  • Docker must be running for local scans; the sandbox image is pulled on first use.
  • A reachable LLM provider is mandatory for local runs. The README recommends GLM-5.3 on Vercel AI Gateway or OpenRouter, and lists other models with caveats about provider access programs.
  • strix view binds to 127.0.0.1 by default and prints a tokened link that grants access to the run; binding to 0.0.0.0 or sharing that link exposes results.
  • The cloud path uploads source code, so strix cloud scans start --source . requires approving the upload.
  • Strix actively attacks its targets. Only run it against systems you own or have explicit written permission to test, and stay inside the agreed scope.
  • The project is Apache-2.0 licensed. Cloud and Enterprise tiers are separate commercial offerings.

HotRepo's take

Strix's trajectory is unusual by any measure: HotRepo's daily snapshots show it climbing from roughly 26.5k stars at the end of June 2026 to about 54.7k by August 17, then 76.2k stars with 8.6k forks on November 17, 2026, when it ranked sixth on the Python daily list. Growth of that shape on a security tool that only appeared in August 2025 points to strong developer interest in agent-driven pentesting rather than to a long track record; no all-language daily rank has been recorded for this repository in HotRepo.

The Show HN thread from August 18, 2025 (102 points, 15 comments) captures the main critique that still matters. tptacek called the project neat but questioned the Xbow comparison, pointing to a fairly basic RCE prompt template discussion and suggesting mature competitors do more than prompts; thegeomaster argued the project looks generated end to end and asked for examples beyond toy problems, noting the screenshot finding would likely be caught by ordinary AI code review comment. In another subthread, commenters joked that some commercial pentesting platforms lean heavily on human staff behind the curtain comment. Those are unverified opinions from early versions, so treat them as evaluation cues rather than conclusions: the question to ask of Strix is whether real PoCs on your own targets hold up.

Written by AI from the project's README, updated . The project's own docs are the source of truth.

Hacker News

Show HN: Strix - Open-source AI hackers for your apps
▲ 102 · 15 comments · August 18, 2025
Show HN: Enter your domain and my open-source agent will hack it
▲ 14 · 3 comments · August 12, 2025
Show HN: Open-source alternative to XBOW
▲ 2 · 0 comments · September 30, 2025

You might also like

  1. KeygraphHQ/shannonAn autonomous agent that reads your web app's source, then runs real exploits against a staging copy to prove vulnerabilities.TypeScript49,899
  2. Armur-Ai/Pentest-Swarm-AIAn open-source, self-hosted alternative to XBOW that runs a swarm of LLM agents to find and prove web/API vulnerabilities.Go3,047
  3. s0ld13rr/pentestcodeMulti-agent autonomous pentesting agent in the terminal with shared engagement state, 18 parser tools, and 13 specialist agents.TypeScript1,080
  4. vxcontrol/pentagiLets security teams run an autonomous, self-hosted multi-agent pentest that drives real tools like nmap and Metasploit against authorized targets.Go32,262
  5. elementalsouls/Claude-BugHunterA Claude Code skill bundle that loads bug-hunting and external red-team methodology by topic, with per-vulnerability-class patterns, validation gates and reporting.Python10,296