For authorized offensive security

The AI agent for Cybersecurity

infosec is a CLI agent for authorized offensive security. It can be used to perform reconnaissance, vulnerability discovery, exploitation, and reverse engineering.

curl -fsSL https://uncensored.chat/infosec.sh | sh

macOS & Linux · standalone binary · no build step

Not a chatbot in a shell

A purpose-built harness that thinks, runs tools, and adapts — the way a real operator works.

Autonomous agent

Give it a goal in plain English — "map the attack surface of this repo" — and it plans, runs tools, reads output, and iterates until the job is done.

Built for offensive security

A harness tuned for authorized recon, vulnerability discovery, and exploitation workflows — not a general chatbot bolted onto a shell.

Uncensored models

Backed by uncensored.chat models that answer real security questions instead of refusing. No hand-holding, no lectures.

Skills & plugins

Extend it with a marketplace of security skills and MCP tools. Install what you need for the engagement; toggle the rest off.

Terminal-native

Lives where you work. Pipe it, script it, run it headless in exec mode, or drive it interactively in the TUI.

Your host, your rules

A single standalone binary. Sign in from the browser, run it on your own box, and keep full control of scope and targets.

infosec — zsh

$ infosec "find hardcoded secrets and weak auth in this repo"

▸ planning · scoping the working tree

→ grep -r "api_key|secret|password" · reading 214 files

! config/services.php:73 — provider key committed in plaintext

! AuthController@login — no rate limit on credential check

▸ verifying · drafting proof-of-concept

✓ 2 confirmed findings · report written to ./infosec-report.md

$

What it can do

The full kill-chain, from one prompt

It doesn't just answer questions — it works the engagement end to end.

Recon & attack surface

Enumerates the codebase, services, endpoints, dependencies, and config to map what an attacker would see.

Vulnerability discovery

Finds injection, auth gaps, secrets, insecure config, and business-logic flaws — then explains the impact.

Exploitation & PoC

Drafts working proofs-of-concept to confirm a finding is real, not a false positive.

Reverse engineering

Deep-reasoning models for binary analysis, deobfuscation, and understanding unfamiliar internals.

Tooling & MCP

Calls real tools and MCP servers — scanners, HTTP clients, your internal APIs — as part of the loop.

Reporting & sessions

Writes structured findings to disk and keeps resumable sessions so long engagements survive restarts.

Model catalog

Pick the right brain for the job

Uncensored, security-tuned models — switch per task with /model or in your config.

Kimi-k3(Latest Coding Model)

Default

Agentic coding, tool use, and security engineering. 1M context.

coding1M contextReasoning

Effort: max

GLM 5.2 (Coding)

General coding, tool use, and security engineering. 1M context.

coding1M contextReasoning

Effort: low · medium · high

The catalog is curated and admin-managed — new models appear here automatically.

Extend it

Bring your own tools

Connect MCP servers, install security skills and plugins from the marketplace, and keep the agent lean with category toggles.

MCP servers

Wire in any Model Context Protocol server — remote (HTTP/SSE) or a local stdio process.

Add a remote MCP server (streamable HTTP/SSE)

Add one that needs a bearer token

Launch a local stdio server

List configured servers

Remove a server

Plugins & skills

Install curated security plugins, or point the CLI at your own git/local marketplace.

List plugins from the curated marketplace

Install a plugin

Add your own marketplace (git or local path)

Refresh a marketplace snapshot

Skills are auto-discovered SKILL.md folders — drop them in ~/.infosec/skills/ or ship them inside a plugin.

Usage that scales with you

Up to 100× more usage

Agentic runs burn tokens fast. Higher plans multiply your monthly pool — measured against Basic, the 1× baseline.

Basic Plan

baseline

1M CLI tokens / month

Included free

Standard Plan

10×vs Basic

10M CLI tokens / month

Needs a paid plan

Best value

Premium Plan

100×vs Basic

100M CLI tokens / month

Needs a paid plan

Questions

What is the infosec CLI?

An autonomous agent that runs in your terminal and carries out security-testing workflows — reconnaissance, vulnerability discovery, exploitation, and reverse engineering — driven by natural-language goals and powered by uncensored models.

Is this only for authorized testing?

Yes. It is intended strictly for authorized offensive-security work: your own systems, CTFs, and engagements where you have explicit permission. You are responsible for staying in scope.

How do I install it?

One command. On macOS and Linux, curl -fsSL the install script and pipe it to sh; on Windows, pipe the PowerShell installer to iex. Either way it drops a standalone binary on your PATH — no Rust toolchain, no Node, no build step. Then run `infosec` and sign in from your browser.

How does usage work?

Every plan includes a monthly pool of CLI tokens. Basic is the baseline (1×); Standard gives 5× that pool and Premium 20×. Usage resets each month, and you can see what is left any time with the /status command or on your dashboard.

Can I add my own tools and models?

Yes. Connect any MCP server (remote or local stdio), install security skills and plugins from the marketplace, and pick from the model catalog per task. Category toggles let you cap tool sprawl on large installs.

Which platforms are supported?

macOS (Apple Silicon and Intel) and Linux (x86_64 and arm64) via the shell installer, and Windows via the PowerShell installer. On Windows you can also run the shell installer inside WSL2.

How is infosec different from PentestGPT and other AI hacking tools?

infosec is a real autonomous agent in your terminal, not a chat wrapper. It runs the full kill-chain end to end — reconnaissance, vulnerability discovery, exploitation, and reverse engineering — connects your own tools over MCP, and runs on uncensored models that answer security questions instead of refusing. Install is a single command, with usage that scales up to 20× on higher plans.

Install in one command

Sign in from the browser, run it on your own host, and start testing in minutes.