8000 GitHub - SomeoneAlt-86/open-mythos: Open source LLM based security agent · GitHub
[go: up one dir, main page]

Skip to content

About

Open source LLM based security agent

Resources

Stars

6 stars

Watchers

1 watching

Forks

Repository files navigation

Open-Mythos

Automated LLM-assisted vulnerability discovery pipeline. Screenshot_20260408_192816

Prerequisites

  • Node.js >= 18
  • An API key in your environment:
    • OPENAI_API_KEY, or
    • OPENROUTER_API_KEY

Installation

npm package under progress soon

npm install -g open-mythos

Quickstart

Run these 3 commands in the repository you want to analyze:

open-mythos recon
open-mythos targets --repo-name "$(basename "$PWD")"
open-mythos research --repo-name "$(basename "$PWD")" --limit 20

What this does:

  • recon creates and stores a threat model.
  • targets creates a deduplicated target queue from that threat model.
  • research scans unscanned targets and stores findings.

Optional dashboard:

open-mythos serve --port 7331

Then open http://localhost:7331.

Full CLI Reference

Global

  • open-mythos --help Show CLI help
  • open-mythos --version Show installed version

run

Run full pipeline (recon -> targets -> research).

open-mythos run [--repo-path <path>] [--repo-name <name>]

Flags:

  • --repo-path <path> Path to target repository (default: current directory)
  • --repo-name <name> Override detected repository name

recon

Generate a threat model for a repository and store it in SQLite.

open-mythos recon [--repo-path <path>]

Flags:

  • --repo-path <path> Path to target repository (default: current directory)

targets

Generate target queue items from an existing threat model.

open-mythos targets [--id <n>] [--repo-name <name>] [--repo-path <path>]

Flags:

  • --id <n> Threat model ID
  • --repo-name <name> Filter by repository name
  • --repo-path <path> Filter by repository path

Note: at least one filter is required (--id, --repo-name, or --repo-path).

research

Scan unscanned targets and store normalized findings.

open-mythos research [--threat-model-id <n>] [--repo-name <name>] [--repo-path <path>] [--limit <n>]

Flags:

  • --threat-model-id <n> Pin research to a specific threat model
  • --repo-name <name> Filter by repository name
  • --repo-path <path> Filter by repository path
  • --limit <n> Max targets pulled per batch (default: 20)

serve

Start read-only dashboard.

open-mythos serve [--port <n>]

Flags:

  • --port <n> Dashboard port (default: 7331)

db

Database utilities.

open-mythos db path
open-mythos db list

Subcommands:

  • db path Print resolved SQLite database path
  • db list List repositories present in database

Environment Variables

Variable Required Default Purpose
OPENAI_API_KEY Conditionally (none) API key for LLM calls (either this or OPENROUTER_API_KEY must be set).
OPENROUTER_API_KEY Conditionally (none) Alternate API key for OpenRouter-compatible endpoint.
OPENAI_BASE_URL No https://openrouter.ai/api/v1 Base URL for OpenAI-compatible API.
DEUS_MODEL No openai/gpt-4.1-mini Model identifier used by the agentic loop.
DEUS_MAX_ITER No 60 Max iterations in the agentic loop.
DEUS_TEMPERATURE No 0.2 Sampling temperature for model calls.
DEUS_APPROVAL_MODE No safe Approval mode passed to loop config.
DEUS_TRUNCATE_THRESHOLD No 40 Message-history truncation threshold.
DEUS_TRUNCATE_KEEP_RECENT No 10 Number of recent turns to keep when truncating.
OPEN_MYTHOS_DB_PATH No src/data/open-mythos.sqlite SQLite file path.
DASHBOARD_PORT No 7331 Port used by dashboard server.

Contributing

  1. Fork and clone the repository.
  2. Install dependencies:
npm install
  1. Set environment variables (copy from .env.example if needed).
  2. Run tests:
npm test
  1. Validate CLI behavior locally:
node bin/open-mythos.js --help
  1. Open a pull request with:
  • clear problem statement
  • focused change set
  • test or validation notes

About

Open source LLM based security agent

Resources

Stars

6 stars

Watchers

1 watching

Forks

Releases

Packages

Contributors

Languages

0