We just created a brand new category for the community: ✨ Show & Tell This is where YOU get to flex what you’re building with Agent Zero: 🔥 Cool workflows ⚡ Automations in action 🚀 Agents doing real work Big or small — doesn’t matter. If you built it, we all want to see it. 👉 Drop your first post in the "✨ Show & Tell" category today and inspire the others!

Hi everyone! I am a cybersecurity professional and have been messing around with using Agent Zero as an automated penetration testing agent since Agent Zero was merged with the hacking edition. I am testing against the intentionally vulnerable OWASP Juice Shop web application which is also running in docker). After learning docker concepts, then Agent Zero concepts, I was able to push Agent Zero pretty far in solving Juice Shop challenges. However, the sequential nature of Juice Shop challenge solving was taking a long time and thus I started looking into using subordinate agents for each challenge (with custom profiles for each, pentestmanager, reconspecialist, pentester). Alas, Agent Zero is not architected to support simultaneous agent swarms at this point. I then pivoted to a custom single agent using the Task Scheduler to create simultaneously executing tasks (using TaskType.AD_HOC, and then run_task with dedicated_context=true) and this appears to be working well so far. I also went to through the journey of attempting to use local LLMs via LM Studio before finally listening to folks here and instead using free LLMs on OpenRouter (currently using Chimera). I still hold hope that I can self-host a ridiculously large opensource LLM, in a private cloud if needed. Abliterated LLMs are particularly useful for ethical hacking and local LLMs are useful for sandboxing cyberlab activities when commercial LLM decline actions. However, there does not appear to be an Agent Zero setting (evironment variable or setting dialog) to adjust LLM timeout value, and slow LLM responses invariably lead to timeout errors. It would be fun to develop the solution to the point where it can target any cyber range system (e.g. infrastructure, edge device) and the agent follows industry penetration testing framework stages and actions from PTES (Penetration Testing Execution Standard), NIST SP 800-115, and OSSTMM, etc.

We're collecting ideas for the chat interface redesign. Please share your ideas or sketches with us, I'll start: My idea is to design the chat like a timeline. Visual structure is ensured via typography and one vertical line. Color of the line is set by the message type. Minimizing and inner scrolling of messages would still work just like now.

Hello Agent Zero Community. This thread is for discussion on local models for Agent Zero. If you are interested and new to that topic, watch the brief beginner videos in preparation for the Local Models for Agent Zero classroom session: https://www.skool.com/agent-zero/classroom/c8000b5f After the live classroom session or if you caught the recording, we would love to hear more about what you are doing with local models. Are you running an enormously big model? Do you have separate local Chat and Utility models? What context sizes are you reaching? We want to hear it all.

Hey all. I have been tinkering around with building a dataset for fine-tuning a local modal and figured I would share it with others who may have interest in doing such a thing. I put it on my github: https://github.com/nafwa03/My-A0-Playground. See output samples (just a few...I have hundreds) Usefulness to this at least for me was to see how the lmf1.2b did not handle summarizing but built a solid tool call. Putting in a persona into the QA was interesting. Using a CodeInjection playbook with qwen3-1.7b just proves that you do not need a huge model to do all of this. While creating synthetic data I found that it may prove useful to those getting stuck in certain situations and to test how your prompting can effect a chain of thought. Small things like using words "Each" vs "Every" make a huge difference. I took @Stirling Goetz as a use case to see if I could make a small qwen3-1.7b model smarter at hacking. I used meta's synthetic data kit feeding it a bunch of hacking books, OWASP guide, some Youtube videos and put Agent 0 persona into the prompt for generating questions and answers, COT, etc.. I used different models to see how the output was different but not that much. I merged LoRA into it and it has more domain knowledge but I don't think that creating a special model for this is necessary (sorry if I am late to this). I will say though that the EMBEDDING may be the secret to all of this but you can also just do MCP I guess. Hope this helps someone who is interested in this stuff. If there is any further interest from any other data nerds I am pretty confident for local models we could do a couple things: 1. Agent profile for Local Use and Cloud 2. Local use comes with tweaks, tips and tricks My contribution :)