Workshops: Agentic Coding

Agentic Coding for DevOps & Infrastructure as Code: Workshop

DevOps teams can use AI agents to write Terraform modules, fix pipelines and analyse incidents faster. Yet an agent with the wrong permissions can damage real infrastructure. This workshop shows how to combine both: high speed within firm safety boundaries.

Who is this workshop for?

  • DevOps and platform engineers
  • Site reliability engineers
  • Cloud architects
  • Teams using Terraform, OpenTofu, Pulumi or Ansible
  • CI/CD owners

What you will take away

  • Have agents write and validate IaC with plan, not apply, as the default.
  • Set permissions, sandboxing and approvals so agents cannot change production resources.
  • Use agents to analyse failed pipelines and maintain CI configuration.
  • Generate Kubernetes manifests and Helm charts and check them against policies.
  • Use agents for incident analysis with logs and metrics as context.

Agenda

  1. Module 1

    Agents and infrastructure

    • Risks: destruction, cost, secrets, drift
    • Claude Code, OpenAI Codex and GitHub Copilot in an infrastructure repository
    • Least privilege and separate roles for agents
  2. Module 2

    Infrastructure as code

    • Exercise: Terraform module with variables, tests and docs
    • Validation with plan, linting and policy as code
    • Importing and refactoring existing resources
  3. Module 3

    CI/CD

    • Having failed pipelines analysed and fixed
    • Creating GitHub Actions or GitLab CI workflows
    • Agents inside the pipeline: headless use with clear limits
  4. Module 4

    Kubernetes and containers

    • Generating manifests and Helm charts
    • Optimising and hardening Dockerfiles
    • Checking against security policies
  5. Module 5

    Operations and incidents

    • Connecting logs, metrics and runbooks via MCP
    • Root cause analysis with the agent
    • Having postmortems drafted

Security first

An agent allowed to run shell commands with admin rights in your cloud is a serious risk. Prompt injection via logs or tickets, misinterpretation or simply an overly broad task can delete resources or run up costs.

We show a layered model: agents work locally and in sandboxes, produce plans and pull requests, and changes to real infrastructure flow only through your existing, reviewed pipelines.

  • Read-only cloud access for analysis
  • No apply without human approval
  • Secrets never in the agent's context
  • Audit logs for all agent actions

Experience from certified operations

jocapps, the company behind neooutput, is ISO 27001 certified. Our AI Solutions Architects and AI Code Engineers combine agentic workflows with the demands of regulated operations.

Format

Hands-on with your repositories in a practice environment, remotely or on-site, in English or German.

Frequently asked questions

Does the agent need access to our cloud?

For most tasks, no. IaC can be written and validated locally. For analysis we recommend read-only roles in non-production environments.

Which clouds and tools do you support?

AWS, Azure, Google Cloud and European providers, with Terraform, OpenTofu, Pulumi, Ansible, Kubernetes and common CI systems. We follow your stack.

Can agents resolve incidents on their own?

We recommend using agents for analysis and proposed fixes. Production changes should go through approved runbooks and people.

Workshop

Get in touch

Want agents to speed up your infrastructure work without putting it at risk?

Tell us a little about your team – we will get back to you within 24 hours with a proposal for content, format and dates.

Prefer to talk?

We only use your details to handle this request. See our privacy policy.