OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Agent Security & Sandboxing

Agent Opfor

Agent Opfor is an open-source developer tool that emulates adversarial attacks against AI agents, LLM applications and MCP servers, then uses an LLM to judge target responses. It offers CLI, browser-extension, IDE, MCP-server and SDK workflows.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Agent Opfor — OpenViST project profile in Agent Security & Sandboxing.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Agent Opfor when you want to exercise an AI system with adversarial tests and inspect the resulting responses, whether through an engineer-led CLI workflow or a browser-based chatbot test.

Who it is for

  • AI and application developers assessing agent or LLM defenses.
  • Security, QA, product and design team members who want to test a deployed chatbot through the browser extension.
  • Teams seeking adversarial testing across agent tools, MCP servers, memory and multi-turn interactions.

Documented uses

  • Run targeted adversarial scans against AI agents and LLM applications.
  • Test MCP servers and agent behavior involving tools, memory and multi-turn reasoning.
  • Review recorded attack prompts, requests, responses and judge verdicts.
  • Use traces from Langfuse or Netra to inform LLM judging of internal agent actions.
Before you choose

Requirements & limits

The project describes itself as open source and identifies Apache 2.0 licensing. The CLI quick start requires a model-provider API key; OpenAI, Gemini and Anthropic are examples.

  • The CLI workflow requires a provider API key and supports a one-shot run or a reusable saved configuration.
  • The browser extension is presented as a no-code route for testing deployed chatbots; separate documentation is linked for its setup.
  • The stated evaluator scope spans OWASP LLM, Agentic AI, MCP and API security suites, plus EU AI Act bias suites.
  • The project says it logs prompts, requests, responses and judge verdicts, and supports trace-aware judging with Langfuse and Netra.
  • Five interfaces are described: CLI, browser extension, MCP server, IDE skills and SDK; each serves a different workflow.

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 and configure the CLI for a reusable scan

Install the CLI, create a saved configuration and run a scan with it.

Before you start

  • Node.js/npm environment for the documented npm installation.
  • An API key for a supported model provider; examples include OpenAI, Gemini and Anthropic.

Steps

  1. Install the CLI globally with npm install -g @keyvaluesystems/agent-opfor-cli.

  2. Set an appropriate provider key in the environment, such as OPENAI_API_KEY.

  3. Run opfor setup to save a configuration under .opfor/configs/.

  4. Run opfor run --config .opfor/configs/ using the saved configuration file.

Expected outcome

A scan runs using a saved configuration that can be reused or committed to CI.

Guide limitations

  • The exact configuration fields and provider-specific setup are not included in this quick-start material.
  • The guide describes the CLI path only, not browser, IDE, MCP-server or SDK setup.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
16 October 2026
Original website title
KeyValueSoftwareSystems/agent-opfor · Official homepage title · retained in its source language
How we review sourcesSuggest a correction