OpenViSTAI · Agents · Blockchain · Crypto
Menu
AI Security & Safety

Ragdrag

Ragdrag is a Python command-line toolkit for authorized security assessment of retrieval-augmented generation pipelines, testing retrieval, generation and ingestion boundaries through phased techniques.

Independent OpenViST project profile, based on linked sources.

Ragdrag — OpenViST project profile in AI Security & Safety.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Ragdrag when you have an authorized RAG endpoint to assess and want to structure checks by phase, with a JSON report for review.

Who it is for

  • RAG application developers assessing their pipeline's security boundaries.
  • Security practitioners conducting authorized tests of RAG endpoints.

Documented uses

  • Assess retrieval and generation behavior with the default R1, R2 and R3 scan phases.
  • Include R6 when the authorized assessment scope covers evasion.
  • Plan ingestion-write assessments with explicit authorization and cleanup controls.
  • Produce JSON reports containing findings, evidence states, capabilities and run status.
Before you choose

Requirements & limits

The project is an open-source repository under the MIT license. The CLI requires Python 3.10 or later.

  • The executable engine registry exposes 20 technique IDs, while the 27-entry taxonomy also includes ideas without executable phase claims.
  • The default scan selects R1, R2 and R3; R6 is opt-in.
  • R4 and R5 can write data and require explicit authorization and cleanup controls.
  • A heuristic finding is classified as inferred and does not by itself prove a vulnerability.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Install Ragdrag from its repository

Set up the CLI and check its reported version.

Before you start

  • Python 3.10 or later
  • A local environment for cloning and installing the repository

Steps

  1. Clone the RAGdrag repository.

  2. Change into the repository directory.

  3. Install the project in editable mode with python -m pip install -e .

  4. Run ragdrag --version.

Expected outcome

The repository's documented setup ends with a CLI version check.

Guide limitations

  • This setup procedure does not assess a RAG endpoint.
Source-based practical guide

Run a default scan against an authorized endpoint

Generate a JSON report from the default R1, R2 and R3 scan.

Before you start

  • Ragdrag installed with Python 3.10 or later
  • An endpoint you own or are explicitly authorized to assess

Steps

  1. Choose the authorized target endpoint.

  2. Run ragdrag scan --target --output report.json.

Expected outcome

The documented command writes a JSON report and uses the default R1, R2 and R3 phases at an active-non-mutating impact ceiling.

Guide limitations

  • The scan sends requests and can expose sensitive responses.
  • This default scan does not include opt-in R6 or ingestion-write phases R4 and R5.
Evidence and transparency

Sources & review

Source record details
Source availability
Active
Sources checked
13 October 2026
Original website title
itsbroken-ai/RAGdrag · Official homepage title · retained in its source language