A curated set of AI tools I actually find useful, plus the pointers that keep the security side honest. The general-purpose stuff up top, the practitioner angle further down. For the deep AI security material (red teaming, MCP security, model scanning, governance), see the Securing AI section on the main Resources page.

A standing caveat, because it matters: don't paste anything personal, confidential, or work-sensitive into a cloud AI you don't control. Most of them log prompts, and many train on them. If the data is sensitive, run a model locally. That principle shapes a lot of what's recommended here.

🤖 Chatbots   🏠 Local & Self-Hosted   🖼️ Image   🎬 Video   🔊 Audio & Voice   💻 Coding   🔎 AI Search & Research   ✍️ Prompting   📊 Benchmarks & Indexes   🛡️ Security Angle

---

🤖 Chatbots

The general-purpose assistants. Most have a usable free tier; sign-up requirements and limits change constantly, so treat specifics as a snapshot.

• Claude — Anthropic's assistant. Strong at writing, reasoning, and code. My daily driver for long-form and technical work.
• ChatGPT — OpenAI's assistant. The one most people mean when they say "AI." Broad tooling and ecosystem.
• Gemini — Google's assistant, with deep integration into the Google ecosystem and a large context window.
• Google AI Studio — Free access to Gemini models with more knobs to turn than the consumer app. Good for prototyping prompts.
• Microsoft Copilot — GPT-backed, wired into the Microsoft 365 world.
• DeepSeek — Capable open-weight models with a generous free tier.
• Qwen — Alibaba's models, strong multilingual and coding performance.
• Mistral — European provider with solid open-weight models and a privacy-forward posture.
• Perplexity — Answer engine that cites sources. More useful than a raw chatbot when you need references.

🏠 Local & Self-Hosted

Where privacy is a concern, this is the answer. Everything here runs on your own hardware, so nothing leaves the box. Directly relevant to a homelab.

• Ollama — The easiest on-ramp to running LLMs locally. One command to pull and run a model. Pairs well with a Proxmox VM.
• LM Studio — Desktop app for discovering, downloading, and chatting with local models. Good GUI for people who don't want to live in a terminal.
• Open WebUI — A self-hosted, ChatGPT-style web front end that sits on top of Ollama or any OpenAI-compatible API. The natural homelab pairing.
• llama.cpp — The inference engine underpinning much of the local-LLM world. Runs quantized models efficiently on CPU or modest GPUs.
• Jan — A fully offline, open-source desktop assistant. Privacy by default.
• GPT4All — Local models with a friendly installer, aimed at running well on everyday hardware.
• AnythingLLM — Self-hosted document chat and RAG over your own files, backed by local or hosted models.
• Can I Run This LLM? — Quick check on whether a given model will fit your hardware before you download 40GB of weights.

🖼️ Image Generation

• ComfyUI — Node-based, fully local Stable Diffusion pipeline. Steep-ish learning curve, near-total control. Runs in the homelab.
• Stability Matrix — A launcher that manages local image-gen frontends and models so you don't fight dependencies.
• Fooocus — Local image generation tuned for good defaults. Closest thing to "just type and get a nice image" while staying self-hosted.
• Gemini / Nano Banana — Google's hosted image generation and editing, if you don't need local.
• Civitai — The big index of Stable Diffusion models and LoRAs. Useful reference even if you generate elsewhere.

🎬 Video Generation

Fast-moving and mostly hosted. Expect sign-ups, credits, and short clips.

• Google Flow — Google's Veo-based video generation, with a daily free allotment.
• Wan AI — Image-to-video with a free daily tier.
• FramePack — A local video-gen option if you have an NVIDIA GPU and want to keep it in-house.

🔊 Audio & Voice

• ElevenLabs — The current benchmark for realistic text-to-speech and voice work.
• Suno — Full song generation from a text prompt. Genuinely impressive, free daily credits.
• Kokoro TTS — A small, high-quality open TTS model you can run locally.
• Applio — Local voice cloning and conversion. Worth understanding, since the same tech underpins voice-based social engineering (see the attacker-tool post).
• Ultimate Vocal Remover — Local stem separation, if you need to split vocals from a track.

💻 Coding Assistants

• Claude Code — Anthropic's agentic CLI coding tool. Strong for multi-file work and refactors.
• Cursor — An AI-first code editor built on VS Code.
• Continue — Open-source IDE assistant that works with local models via Ollama, keeping code on your machine.
• aider — A terminal-based pair programmer that edits your git repo directly, model-agnostic.

🔎 AI Search & Research

General search-style assistants that cite sources, plus the more academic, paper-focused research bots.

Search & answer engines

• Perplexity — Cited answers, good for quick research where provenance matters.
• Exa — A search API and engine built for AI-style semantic queries.
• Google AI Mode — Google's AI-answer search mode, with a free tier.
• Ask Brave — Brave's AI answer engine, privacy-forward.

Document & literature research

• NotebookLM — Upload your own sources and query them. Solid for working through docs, papers, or a pile of PDFs. Can also generate audio overviews.
• Elicit — Research assistant aimed at working through academic literature and systematic reviews.
• SciSpace — Chat with papers, get explanations of dense passages, and search across a large literature corpus.
• Alphaxiv — A research chatbot layered over arXiv, for interrogating preprints.
• Bohrium — A research-focused chatbot aimed at scientific work.
• Scinito — Another research assistant for querying academic sources.
• Sci-Bot — A Sci-Hub-backed research chatbot.
• Consensus — Search engine that surfaces findings from peer-reviewed research and summarizes what the literature actually says.
• PrivateGPT — Self-hosted document chat, so your source material never leaves your machine. Fits the homelab and the privacy principle up top.

General assistants worth using for research

• Claude — Strong at reasoning over long documents and synthesizing across sources.
• ChatGPT — Deep Research mode does multi-step, cited investigations.
• Gemini — Large context window and its own Deep Research mode.

✍️ Prompting

• Prompt Engineering Guide — The most thorough open guide to prompting, including adversarial and injection techniques worth understanding defensively.
• Learn Prompting — Structured free courses, including a prompt-hacking track.
• Anthropic's Prompt Engineering docs — Practical, model-specific guidance that generalizes well.

📊 Benchmarks & Indexes

For when you need to pick a model and want more than vibes.

• Artificial Analysis — Independent benchmarks for chat, image, and video models, plus pricing and speed.
• LM Arena — Crowd-voted head-to-head model rankings.
• Models.dev — A model database with capabilities and pricing.
• LLM Pricing — Quick API cost comparison across providers.
• FutureTools — A broad directory of AI tools, filterable by free options.

---

🛡️ Security Angle

The reason this page reads differently from a generic AI list. A few tools and references that sit at the intersection of "AI tool" and "security practitioner," pulled to the top so they're easy to find. The full treatment lives in the Securing AI section of the main Resources page.

• Ollama + Open WebUI — A private, self-hosted assistant stack. The single most useful setup for anyone who wants AI help without shipping sensitive data to a third party.
• Gandalf — Lakera's browser-based prompt-injection game. The gentlest possible intro to why LLM input is dangerous.
• Garak — NVIDIA's LLM vulnerability scanner. Point it at a model and see what falls out.
• PyRIT — Microsoft's risk-identification toolkit for generative AI. For actually red-teaming a deployment.
• OWASP Top 10 for LLMs — The starting point for reasoning about LLM application risk.
• MITRE ATLAS — The ATT&CK-style knowledge base for adversarial ML. Where to map AI threats.
• Embrace The Red — Johann Rehberger's blog, some of the best hands-on LLM and agent attack research going.

For the deep material (AI red teaming, blue teaming, MCP and agent security, model scanning, governance frameworks, vulnerable-by-design labs) head to the Securing AI section on the Resources page. That's where I keep the serious security tooling.