Skip to content
ML2

Shannon: open-source A.I. pentester for web apps and APIs

Added Copyable prompt below

Published Last verified Released

Our verdict

v3.5.0 (27 Oct 2026) is a maintenance release: it repairs shannon stop, polishes the PDF report cover and adds an experimental --exp-php2shell probe that checks whether a PHP server is patched against the php2shell crash (it deliberately crashes one PHP worker, so staging only). The core is why it is worth a look: Shannon combines source-code analysis with live exploitation against a staging or local copy of your app, and drops any finding it can't prove with a working exploit, so the report is short and evidence-backed rather than a scanner dump. It is widely watched (about 50,300 GitHub stars as of 2026-10-29, gaining roughly 476 a day in our scan that day), but stars are not a quality signal; Keygraph says findings still need human review. Use it only on systems you own or have written permission to test.

Who it's for
Developers and small teams who ship web apps or APIs, have the source code and a staging environment, and want exploit-backed security testing between human pentests
Leverage
An exploit-verified pentest of your own app from one npx command, repeatable in CI on every release
Effort
Medium
Date added

What it does

Shannon (KeygraphHQ/shannon, AGPL-3.0, npm @keygraph/shannon) is an autonomous A.I. pentester for web applications and their APIs. It maps the app's architecture, trust boundaries, interfaces and data flows from the source, explores the running app, runs specialised agents for injection, XSS, SSRF, broken authentication and broken authorization, merges and deduplicates candidates, then has exploitation agents attempt real proof-of-concept attacks through browser automation and command-line tools. Anything it cannot demonstrate is discarded. Output is PDF and Markdown reports with evidence, plus JSON and SARIF 2.1.0. It supports authenticated testing (login flows, TOTP, email auth), rules of engagement, resumable workspaces, and an official GitHub Action and GitLab CI/CD component that can fail a pipeline on exploited findings above a severity threshold. It needs Docker and Node.js 18+, runs each scan in an ephemeral container with the repo mounted read-only, and uses your own model key (Anthropic, OpenAI, xAI, AWS Bedrock, gateways, or local models via Ollama, vLLM or LM Studio); Keygraph says it never proxies model traffic. It is white-box only: black-box testing, SCA, secrets scanning and fix PRs are in Keygraph's paid Pro and Enterprise editions. Risks: it runs real attacks that can mutate data, so use disposable staging data (Keygraph suggests a throwaway VM); each run costs model tokens on your bill; reading untrusted repos exposes it to prompt injection; and AGPL-3.0 means that if you modify Shannon and let others use it over a network you must publish your changes (Keygraph sells commercial licences for other terms).

How builders use it

  • Clear your model provider's cyber-security safeguards first (Anthropic and OpenAI can interrupt security workloads mid-run), then run npx @keygraph/shannon@latest setup to store credentials.
  • Point it at a staging or local copy you own: npx @keygraph/shannon@latest start -u <staging URL> -r <path to repo>. Never production; exploits can create users and change or delete data.
  • Read the PDF or Markdown report, reproduce each proof of concept yourself, then fix and re-run (interrupted scans resume without repeating finished work).
  • Once a manual run looks sane, add KeygraphHQ/shannon-action@v1 to a private repo's workflow with fail-on-severity and upload-sarif to gate releases.

What to expect

A report listing only vulnerabilities Shannon actually exploited on your staging app, with reproduction evidence and SARIF for code scanning, in roughly 1 to 1.5 hours per full run, plus your own model API costs.

Copy this: Install + first authorized pentest

# Official source: github.com/KeygraphHQ/shannon (AGPL-3.0)
# Only test apps you own or have explicit written permission to test. Not production.
# Needs Docker, Node.js 18+ and your own model API key.

# Interactive launcher (guided setup and first pentest)
npx @keygraph/shannon@latest

# Or step by step:
npx @keygraph/shannon@latest setup
npx @keygraph/shannon@latest start \
  -u https://staging.your-app.com \
  -r /path/to/your/repo

# v3.5.0 experimental php2shell patch check (crashes one PHP worker; staging only)
npx @keygraph/shannon start --exp-php2shell -u https://staging.example.com

Related discoveries

See the newest picks →
  • ToolCoding & Shipping Agents

    Sphinx: quiz yourself before shipping agent code

    MIT Git hook that quizzes you on what your coding agent's change does before a protected push continues.

    Effort: MediumPrompt#Plugin / extension#Open source

  • ToolCoding & Shipping Agents

    Cloudflare security-audit skill: agent bug hunts

    Cloudflare's MIT skill that turns Claude Code, Codex or Cursor into a six-phase security auditor where a fresh agent tries to disprove every finding.

    Effort: MediumPrompt#Plugin / extension#Open source

  • WorkflowCoding & Shipping Agents

    Auto lint-and-test fix loop with Aider

    Make the agent run your linter and tests after each edit and fix what breaks.

    Effort: LowPrompt

Get the best new A.I. picks in your inbox

The best new agent workflows and tools we’ve vetted, with a clear verdict on each. Short, occasional, useful.

More in Coding & Shipping Agents →

Free. No spam. Unsubscribe anytime.