OpenClaw: Understanding and Deploying Agentic AI

“OpenClaw: Understanding and Deploying Agentic AI” is a hands-on 1.5-day training focused on OpenClaw, an open-source AI agent widely adopted across the tech ecosystem. Its aim is to give participants a clear understanding of the tool together with the practical skills to install, configure and use it on a daily basis in a secure way. No technical prerequisites are required.

The first part (2h) lays the foundations of agentic AI and introduces OpenClaw (origins of the project, positioning, use cases). It covers the key concepts (LLMs, MCP, skills, Gateway, Agent Runtime), the main models on the market (GPT, Claude, Gemini, open-source models) and the available integrations (email, Slack, Discord, WhatsApp).

The second part (2h) is dedicated to the secure deployment of OpenClaw. Like any powerful tool, OpenClaw requires knowing the right practices to use it safely. Participants install and configure their own environment, learn how to secure their access, how to audit skills before installation and how to build the right habits. Each participant ends up with a working and secured OpenClaw environment.

The third part (2h) puts participants to work on a concrete case: building, with OpenClaw, an agent that monitors information sources on a chosen topic, synthesises the relevant content and generates a newsletter ready to be sent by email. The fourth part (6h) is dedicated to each participant’s personal project, interspersed with sessions on advanced techniques (SKILL.md, MCP servers, cron scheduling).

This training offers a complete journey: from theoretical fundamentals to the wrap-up of a personal project, going through the secure installation of OpenClaw and the building of a monitoring agent that produces a newsletter. Participants come out autonomous and responsible in their daily use of OpenClaw.

Theme: Agentic AI Fundamentals 2 hours  

Objective / Output

• Understand the agentic AI ecosystem and where OpenClaw fits in
• Master the key concepts: LLMs, MCP, skills, autonomous agents
• Know how to choose the right AI model depending on the use case
• Understand OpenClaw’s architecture and its capabilities
• Identify concrete use cases in one’s professional context

Theme: Secure setup and hands-on with OpenClaw  2 hours

Objective / Output

• Know how to install and configure OpenClaw securely
• Be aware of security risks and know how to protect against them
• Master security best practices (credentials, permissions, skills auditing)
• Be able to create and test a first OpenClaw agent

Content

• Installing OpenClaw and configuring the openclaw.json file
• Securing the environment: managing credentials, permissions and network
• Security best practices: protection against known vulnerabilities and prompt injections
• Skills auditing: checking a skill before installation, avoiding common pitfalls
• Configuring tool policies: sandbox, allowlists/denylists, restricting shell commands
• First working agent: connecting to a messaging channel (WhatsApp or Discord) and live testing

Theme: Monitoring & newsletter generation with OpenClaw  2 hours

Objective / Output

• Know how to build a topic-based monitoring agent end to end
• Master the workflow with OpenClaw (prompting, iteration, feedback)
• Know how to produce a usable newsletter from web sources
• Understand how to connect OpenClaw to email (and, as an optional bonus, to Slack and WhatsApp)
• Become more efficient when interacting with an AI agent

Content

• Case briefing: topic-based monitoring agent and generation of a newsletter sent by email
• Information gathering and synthesis: web search, source filtering, summarisation, perspective
• Newsletter generation and formatting via dedicated skills (structure, editorial tone, proofreading)
• Delivery by email; optional bonus: Slack and WhatsApp. The email, Slack and WhatsApp accounts are provided ahead of the training to avoid any time lost on configuration.
• Methodology: effective prompting, fast iteration, managing conversational memory
• Collective debrief and sharing of the newsletters produced by each participant

Theme: Personal project, advanced techniques & final wrap-up 6 hours

Objective / Output

• Deliver a complete personal project with OpenClaw
• Master advanced techniques (custom skills, MCP, automation)
• Know how to create custom skills and extend OpenClaw’s capabilities
• Present the project to the group as a demonstration and feedback session

Content

• Scoping and framing of the personal project (or extending the monitoring / newsletter case)
• Tips & best practices: writing SKILL.md, configuring MCP in openclaw.json, security
• Advanced techniques: cron scheduling, persistent memory, parallel multi-agents, local models
• One-on-one coaching by the trainers and real-time troubleshooting

• Understand OpenClaw’s architecture (Gateway, Agent Runtime, skills) and the key concepts of agentic AI (LLMs, MCP)
• Know the main AI models on the market (GPT, Claude, Gemini, open-source models) and the available integrations (WhatsApp, Discord, Slack, email)
• Deploy OpenClaw securely by applying best practices (credentials, permissions, skills auditing)
• Identify the security risks linked to AI agents and know how to protect against them
• Create agents, custom skills (SKILL.md) and configure MCP servers
• Deliver a complete project with OpenClaw, from scoping to the final wrap-up
• Assess the security and governance considerations for integrating OpenClaw into production in an organisation