AgentKit is a free, open-source Python toolkit that helps beginners diagnose their PC for local LLM support, pick the right model, and chat from the terminal.

Do I need a powerful PC to use AgentKit?

It depends on the model. AgentKit diagnoses your hardware and recommends models that will run on your system — from 4GB RAM upwards.

Yes. AgentKit is MIT-licensed open source — free to use, modify, and distribute.

I built AgentKit — a friendly AI agent toolkit (open source)

AI agents are powerful, but getting started is unnecessarily hard. The guides assume you already know what a context window is, the tooling is fragmented across a dozen different apps, and if you're new to all this — or you just want to know whether your laptop can run a local model — you're left guessing.

AgentKit is my attempt to fix that. It's a free, open-source Python toolkit that runs in your terminal and helps anyone — from complete beginner to experienced developer — evaluate their setup, get configured, and start extracting real value from AI.

It's on GitHub, MIT-licensed, and takes about 30 seconds to install.

ⓘ Sponsored: This post is from the team behind LawImNotSure, a free UK legal document generator. We built AgentKit as an open-source side project — it's not a product, there's no paywall, and we don't track you.

What problem does it solve?

I kept running into the same wall: people hear "AI agents can do X, Y, Z" and want to try, but they don't know:

Whether their computer can even run a local model
Which tool to install (Ollama? LM Studio? llama.cpp?)
What model size their hardware can handle
How to get past the "OK but what do I type next?" stage

AgentKit answers those questions in one place, with a friendly terminal interface.

What it does

diagnose Check if your system can run local LLMs

Scans your GPU, RAM, CPU, and existing AI tooling. Tells you honestly whether local inference is viable, what model sizes you can handle, and whether cloud APIs are a better bet.

Example output on this machine:

  GPU: NVIDIA GeForce RTX 3080 Laptop GPU
  VRAM: 16.0 GB
  CUDA: 12.4
  Score: 100/100 — Local LLM: YES
  Recommended: Qwen2.5-7B, Llama-3.1-8B, Mistral-7B

wizard Step-by-step beginner setup

An interactive wizard that asks what you want to do, what experience you have, and walks you through the right setup. No prior knowledge assumed. It'll help you install Ollama, pick a model, set up an API key — whatever matches your situation.

chat Terminal chat with any model

Talk to Ollama, OpenAI, Anthropic, Groq, LM Studio, or any CLI-based model from one interface. Sessions save to disk. Built-in tools let the AI read files and run shell commands (agent mode).

recommend Model recommendations for your goal

Tell it you want to code, write, learn, or just chat — it suggests the right models and explains why. Recommendations adjust to your hardware.

Quick start

Install from the repo:

git clone https://github.com/ellistvizor-cloud/agentkit
cd agentkit
pip install -e .

Then:

# Check your system
agentkit diagnose

# Run the beginner wizard (interactive)
agentkit wizard

# Start chatting (uses your default provider)
agentkit chat

# Chat with a specific model
agentkit chat --provider ollama --model llama3.2
agentkit chat --provider groq
agentkit chat --provider openai --model gpt-4o-mini

That's it. No config files to edit, no API endpoints to memorise. If Ollama is running on your machine, agentkit chat just works.

Why I built it

The team behind LawImNotSure builds free tools that help ordinary people handle everyday legal problems without paying hundreds in solicitor fees. We use AI extensively in our own products, and we kept thinking: everyone's talking about AI agents, but the on-ramp is terrible for normal people.

So we built the on-ramp we wished existed. AgentKit is our contribution back to the community — a toolkit that makes AI approachable, not something only developers with GPU clusters can use.

Who it's for

Complete beginners who want to try AI but don't know where to start
People with a GPU who want to run local models but aren't sure what their hardware can do
Developers who want a simple, scriptable CLI for talking to multiple providers
Anyone tired of 10 different apps and wants one tool that does it all

Supported providers

Ollama — local, easiest to set up
LM Studio — local, GUI option
OpenAI — cloud, OPENAI_API_KEY
Anthropic — cloud, ANTHROPIC_API_KEY
Groq — cloud, fast and free for some models
Together AI — cloud
CLI (any) — pipe prompts to any command-line model runner

What's next

This is v1.0. There's plenty to add — more provider backends, more built-in tools (web search, code execution), better beginner explanations in the wizard, and proper tests.

The best way to help is to try it, tell us what's confusing, and submit a PR. The code is small and readable — about 3,500 lines across 8 modules.

💻 View on GitHub 📄 Browse all free tools

AgentKit is free and open source (MIT). LawImNotSure is a free UK legal document generator — your first document is free. If you found this useful, consider browsing our other free tools or starring the repo.