Responsible AI Agent Workshop: Build AI agents with your team that you can trust.

Q: What are open-weights AI models?

Open weights refers to AI models whose trained model parameters are publicly available for download. You can run them on your own hardware or with the provider of your choice, fine-tune them, and use them without depending on a single cloud vendor. Unlike closed models such as GPT-5.5 or Claude Opus 4.7, which only run through the vendor's API, your data and workloads stay under your control. Examples: Kimi K2.6, Mistral, DeepSeek.

Q: What is the difference between open source and open weights?

Open source means a model's source code is publicly available: training code, architecture and ideally also the training data. Anyone can rebuild the model from scratch, audit it and adapt it. Open weights only means the trained model parameters are available for download. How the model was trained, on what data and with what code, often stays closed. In practice: with open weights you can host and run the model yourself right away. With open source you can additionally retrace it from scratch, replicate it and retrain it on different data. Most "open" AI models such as Kimi K2.6, Mistral and DeepSeek are strictly speaking open weights, not open source. They release the weights but not the full training pipelines.

Responsible AI Agent Workshop with N3XTCODER

What to expect

Do you want AI agents for your team that run on open-source technology, keep your data with you, know your processes, and still deliver top performance? In this workshop, every team member builds their own agent. Step by step. With our support.

The workshop is for teams from startups, NGOs and SMEs.

The tech stack

We prepare the tech stack ahead of the workshop so we can dive straight in.

Hermes Agent

Currently the most-used open-source agent.

Kimi K2.6

The smartest open-weights model, cost-efficient to host in Europe.

Infomaniak

European infrastructure, open-source data centre in Geneva, 100% renewable energy.

Hermes Agent

Hermes is the agent that grows with you. Developed by Nous Research, it runs on your own server, learns from every interaction and stores its knowledge in auto-generated skills. You start a task on Telegram and continue it on Slack. Hermes remembers everything. You can run Hermes locally on your computer or on your own server, secure and isolated. With currently 157,000 GitHub stars, Hermes is one of the most popular AI agents worldwide and even overtook OpenClaw in usage on OpenRouter in May. Hermes is completely model-agnostic. You can connect any number of LLM providers simultaneously and switch between them flexibly.

Persistent memory: the agent never forgets how it solved a problem
Multi-platform: chat with your agent via Signal, Matrix, Mattermost, Telegram, Discord, Slack, WhatsApp, email or CLI
Source code: github.com/NousResearch/hermes-agent

By playing the video, you consent to the transfer of data to the respective video platform (e.g., YouTube, Vimeo).

Kimi K2.6

Kimi is particularly well-suited for AI agents. The open-weights model is developed by the Chinese startup Moonshot AI. In the Artificial Analysis Intelligence Index, Kimi K2.6 ranks second with 57 points, alongside Claude Opus 4.7 and Gemini 3.1 Pro, just behind GPT 5.5.

Benchmark: artificialanalysis.ai/models
Download open weights and use them. No vendor lock-in, self-hosting possible
Significantly cheaper than Claude or GPT. Hosted on Infomaniak: €0.60 / 1M input tokens, €3.00 / 1M output tokens

By playing the video, you consent to the transfer of data to the respective video platform (e.g., YouTube, Vimeo).

Infomaniak

Infomaniak is the European alternative to US cloud giants. Founded in 1990 as an IT club in Geneva, Infomaniak is now Switzerland's largest web hosting company, with over one million users and 200 employees (70% in R&D). The company is B Corp certified and has been committed to sustainability since 2007, running on 100% renewable energy since 2014.

Own solar power plants: Since 2024, two installations with 1,085 Meyer Burger modules (420 kWp) producing up to 400,000 kWh per year
Independent foundation: The Infomaniak foundation holds the majority of voting rights. No investor can change its independence.
D4 datacentre: Heats 6,000 Geneva households in winter. In summer, 20,000 people enjoy a 5-minute daily shower with recovered heat.

By playing the video, you consent to the transfer of data to the respective video platform (e.g., YouTube, Vimeo).

No vendor lock-in

You can swap out any component later: LLM, agent, hosting. You have full sovereignty.

Process

1. Attend the free webinar

Your team can prepare by joining our free webinar State of Responsible AI.

2. Preparation

N3XTCODER deploys a Hermes Agent on Infomaniak infrastructure for every workshop participant and connects it to Kimi K2.6. Ready to use on workshop day. Every team member starts with their own instance.

3. Workshop

Setup and introduction to the open-source AI stack, personalisation with your own data. Each participant builds their own skill for their use case.

4. Keep using your agents

After the workshop we can shut down the server with the agents, or your team takes them over and develops them further. The N3XTCODER team can support you with that on request.

What are skills?

Skills are reusable instructions for specific tasks. They work like recipes that agents execute step by step. Every skill is an easy-to-read SKILL.md file that you can edit with any text editor.

What does a skill define?

A skill defines:

Who: The role (e.g. "Instagram content expert")
What: The task (e.g. "Create a LinkedIn post")
How: Structure, writing style and formatting
Quality: Criteria and control questions

Examples

create-instagram-post: posts with hook, hashtags and image prompt
extract-insights: filter and structure insights from texts
check-brand: ensure brand compliance

What the research shows

〰️

N3XTCODER is part of the Civic Coding Initiative run by the German federal ministries BMAS, BMBFSFJ, BMUKN and BMDS to advance AI for the common good. In May 2026, Civic Coding published a thesis paper. The findings support the design of this workshop.

Thesis: Experience is the scarcest resource

"The greatest learning effect comes not from knowledge transfer, but from working together on concrete problems."

That is exactly what the workshop is built for: each team member builds their own skill for their own use case.

Thesis: Strategic AI competence is the blind spot

"The decisive competence need is not in handling tools, but in the strategic positioning of AI as a transformation factor."

Instead of just trying out tools, you take over a complete open-source stack with full sovereignty.

Thesis: Shared structures are the lever

"The central lever for competence building lies in building lasting, shared structures for exchange and collective learning."

The N3XTCODER team can support your agent development beyond the workshop on request.

This workshop is right for your team if:

✅ you want to learn hands-on, not just talk about AI

✅ you want to use open-source technology instead of being tied to a US cloud provider

✅ you want to host your data in Europe, on renewable energy

Book a call with Simon about your team and your challenges

We will reply with a proposal and a date, usually within a working day.

Simon Stegemann
Co-Founder and CEO

Pick a date for your video call

Have a 20-minute video call with our co-founder and CEO Simon

Receive a tailored offer

Why N3XTCODER is the leading AI agency in Germany

We at N3XTCODER have crafted digital solutions for global challenges with top partners across tech, civil society, and the public sector for over 10 years.

We have worked with over 160 organisations to get their AI solutions implemented.
Over 100,000 people have learnt how to use AI for the common good through our online video course AI for Impact.

Our mission: Tech with purpose, tech for good.

With our projects, we aim to contribute to the 17 Global Goals of the United Nations. Example: we can host your AI solutions on 100% renewable energy in Europe and choose data centres that heat thousands of households through district heating.

Webinar: State of Responsible AI

Friday, 5 June 2026, 12:30 - 14:00 CEST, online. Language: German.

A free overview of the most pressing challenges and opportunities of artificial intelligence. We provide orientation and show concretely how to use AI responsibly without compromising on quality.

Learn more and register

Frequently asked questions

What are open-weights AI models?

What is the difference between open source and open weights?