rogue

Name: rogue
Author: qualifire-dev

byqualifire-dev· 1 MCP server

Stars

1.1k

Forks

164

How to install rogue?

qualifire-dev/rogue

$git clone https://github.com/qualifire-dev/rogue

Installs into the current project.

›Prefer a prompt? Paste this to your agent

Install & use

Install rogue by running `git clone https://github.com/qualifire-dev/rogue`, then use it for the current task and follow its documentation at https://github.com/qualifire-dev/rogue.

Files · 1

View on GitHub

README.md

1	# Rogue — AI Agent Evaluator & Red Team Platform
2
3	![](https://pixel.qualifire.ai/api/record/rogue)
4
5	<div align="center">
6
7	<a href="https://trendshift.io/repositories/15191" target="_blank"><img src="https://trendshift.io/api/badge/repositories/15191" alt="qualifire-dev%2Frogue \| Trendshift" style="width: 250px; height: 55px;" width="250" height="55"/></a>
8
9	![Tests](https://github.com/qualifire-dev/rogue/actions/workflows/test.yml/badge.svg?branch=main)
10
11	<img src="./freddy-rogue.png" width="200"/>
12
13	Stress-test your AI agents before attackers do.
14
15	[Discord Community](https://discord.gg/EUfAt7ZDeK) · [Quick Start](#-quick-start) · [Documentation](./docs/)
16
17	</div>
18
19	---
20
21	## Two Ways to Harden Your Agent
22
23	<table>
24	<tr>
25	<td width="50%" valign="top">
26
27	### 🎯 Automatic Evaluation
28
29	Test your agent against business policies and expected behaviors.
30
31	- Define scenarios & expected outcomes
32	- Verify compliance with business rules
33	- Watch live conversations as Rogue probes your agent
34	- Get detailed pass/fail reports with reasoning
35
36	Best for: Regression testing, behavior validation, policy compliance
37
38	</td>
39	<td width="50%" valign="top">
40
41	### 🔴 Red Teaming
42
43	Simulate adversarial attacks to find security vulnerabilities.
44
45	- 75+ vulnerabilities across 12 security categories
46	- 20 attack techniques (encoding, social engineering, injection)
47	- CVSS-based risk scoring
48	- 8 compliance frameworks (OWASP, MITRE, NIST, GDPR, EU AI Act)
49
50	Best for: Security audits, penetration testing, compliance reporting
51
52	</td>
53	</tr>
54	</table>
55
56	---
57
58	## Architecture
59
60	Rogue operates on a client-server architecture with multiple interfaces:
61
62	\| Component \| Description \|
63	\| ---------- \| ------------------------------------------- \|
64	\| Server \| Core evaluation & red team logic \|
65	\| TUI \| Modern terminal interface (Go + Bubble Tea) \|
66	\| CLI \| Non-interactive mode for CI/CD pipelines \|
67
68	https://github.com/user-attachments/assets/b5c04772-6916-4aab-825b-6a7476d77787
69
70	### Supported Protocols
71
72	\| Protocol \| Transport \| Description \|
73	\| ---------- \| -------------------- \| ---------------------------------------------------------------------------------- \|
74	\| A2A \| HTTP \| [Google's Agent-to-Agent](https://a2a-protocol.org/latest/) protocol \|
75	\| MCP \| SSE, STREAMABLE_HTTP \| [Model Context Protocol](https://modelcontextprotocol.io/) via `send_message` tool \|
76	\| Python \| — \| Direct Python function calls (no network protocol) \|
77
78	See examples in [`examples/`](./examples/) for reference implementations.
79
80	#### Python Entrypoint
81
82	For agents implemented as Python functions without A2A or MCP:
83
84	1. Create a Python file with a `call_agent` function:
85
86	```python
87	def call_agent(messages: list[dict]) -> str:
88	"""
89	Process conversation and return response.
90
91	Args:
92	messages: List of {"role": "user"\|"assistant", "content": "..."}
93
94	Returns:
95	Agent's response as a string
96	"""
97	# Your agent logic here
98	latest = messages[-1]["content"]
99	return f"Response to: {latest}"
100	```
101
102	2. Run Rogue with Python protocol:
103
104	```bash
105	uvx rogue-ai cli \
106	--protocol python \
107	--python-entrypoint-file ./my_agent.py \
108	--judge-llm openai/gpt-4o-mini
109	```
110
111	Or via TUI: select "Python" as the protocol and enter the file path.
112
113	See [`examples/python_entrypoint_stub.py`](./examples/python_entrypoint_stub.py) for a complete example.
114
115	---
116
117	## 🔥 Quick Start
118
119	### Prerequisites
120
121	- `uvx` — [Install uv](https://docs.astral.sh/uv/getting-started/installation/)
122	- Python 3.10+
123	- LLM API key (OpenAI, Anthropic, or Google)
124
125	### Installation
126
127	```bash
128	# TUI (recommended)
129	uvx rogue-ai
130
131	# CLI / CI/CD
132	uvx rogue-ai cli
133	```
134
135	### Try It With the Example Agent
136
137	```bash
138	# All-in-one: starts both Rogue and a sample T-shirt store agent
139	uvx rogue-ai --example=t

6 picks

Type

rogue

byqualifire-dev· 1 MCP server

Stars

1.1k

Forks

164

How to install rogue?

qualifire-dev/rogue

$git clone https://github.com/qualifire-dev/rogue

Installs into the current project.

›Prefer a prompt? Paste this to your agent

Install & use

Install rogue by running `git clone https://github.com/qualifire-dev/rogue`, then use it for the current task and follow its documentation at https://github.com/qualifire-dev/rogue.

Files · 1

View on GitHub

README.md

1	# Rogue — AI Agent Evaluator & Red Team Platform
2
3	![](https://pixel.qualifire.ai/api/record/rogue)
4
5	<div align="center">
6
7	<a href="https://trendshift.io/repositories/15191" target="_blank"><img src="https://trendshift.io/api/badge/repositories/15191" alt="qualifire-dev%2Frogue \| Trendshift" style="width: 250px; height: 55px;" width="250" height="55"/></a>
8
9	![Tests](https://github.com/qualifire-dev/rogue/actions/workflows/test.yml/badge.svg?branch=main)
10
11	<img src="./freddy-rogue.png" width="200"/>
12
13	Stress-test your AI agents before attackers do.
14
15	[Discord Community](https://discord.gg/EUfAt7ZDeK) · [Quick Start](#-quick-start) · [Documentation](./docs/)
16
17	</div>
18
19	---
20
21	## Two Ways to Harden Your Agent
22
23	<table>
24	<tr>
25	<td width="50%" valign="top">
26
27	### 🎯 Automatic Evaluation
28
29	Test your agent against business policies and expected behaviors.
30
31	- Define scenarios & expected outcomes
32	- Verify compliance with business rules
33	- Watch live conversations as Rogue probes your agent
34	- Get detailed pass/fail reports with reasoning
35
36	Best for: Regression testing, behavior validation, policy compliance
37
38	</td>
39	<td width="50%" valign="top">
40
41	### 🔴 Red Teaming
42
43	Simulate adversarial attacks to find security vulnerabilities.
44
45	- 75+ vulnerabilities across 12 security categories
46	- 20 attack techniques (encoding, social engineering, injection)
47	- CVSS-based risk scoring
48	- 8 compliance frameworks (OWASP, MITRE, NIST, GDPR, EU AI Act)
49
50	Best for: Security audits, penetration testing, compliance reporting
51
52	</td>
53	</tr>
54	</table>
55
56	---
57
58	## Architecture
59
60	Rogue operates on a client-server architecture with multiple interfaces:
61
62	\| Component \| Description \|
63	\| ---------- \| ------------------------------------------- \|
64	\| Server \| Core evaluation & red team logic \|
65	\| TUI \| Modern terminal interface (Go + Bubble Tea) \|
66	\| CLI \| Non-interactive mode for CI/CD pipelines \|
67
68	https://github.com/user-attachments/assets/b5c04772-6916-4aab-825b-6a7476d77787
69
70	### Supported Protocols
71
72	\| Protocol \| Transport \| Description \|
73	\| ---------- \| -------------------- \| ---------------------------------------------------------------------------------- \|
74	\| A2A \| HTTP \| [Google's Agent-to-Agent](https://a2a-protocol.org/latest/) protocol \|
75	\| MCP \| SSE, STREAMABLE_HTTP \| [Model Context Protocol](https://modelcontextprotocol.io/) via `send_message` tool \|
76	\| Python \| — \| Direct Python function calls (no network protocol) \|
77
78	See examples in [`examples/`](./examples/) for reference implementations.
79
80	#### Python Entrypoint
81
82	For agents implemented as Python functions without A2A or MCP:
83
84	1. Create a Python file with a `call_agent` function:
85
86	```python
87	def call_agent(messages: list[dict]) -> str:
88	"""
89	Process conversation and return response.
90
91	Args:
92	messages: List of {"role": "user"\|"assistant", "content": "..."}
93
94	Returns:
95	Agent's response as a string
96	"""
97	# Your agent logic here
98	latest = messages[-1]["content"]
99	return f"Response to: {latest}"
100	```
101
102	2. Run Rogue with Python protocol:
103
104	```bash
105	uvx rogue-ai cli \
106	--protocol python \
107	--python-entrypoint-file ./my_agent.py \
108	--judge-llm openai/gpt-4o-mini
109	```
110
111	Or via TUI: select "Python" as the protocol and enter the file path.
112
113	See [`examples/python_entrypoint_stub.py`](./examples/python_entrypoint_stub.py) for a complete example.
114
115	---
116
117	## 🔥 Quick Start
118
119	### Prerequisites
120
121	- `uvx` — [Install uv](https://docs.astral.sh/uv/getting-started/installation/)
122	- Python 3.10+
123	- LLM API key (OpenAI, Anthropic, or Google)
124
125	### Installation
126
127	```bash
128	# TUI (recommended)
129	uvx rogue-ai
130
131	# CLI / CI/CD
132	uvx rogue-ai cli
133	```
134
135	### Try It With the Example Agent
136
137	```bash
138	# All-in-one: starts both Rogue and a sample T-shirt store agent
139	uvx rogue-ai --example=t

6 picks

Type

rogue

How to install rogue?

Related

rogue

How to install rogue?

Related