Contents
An agentic coding workshop for DevOps and platform engineering tackles two questions at once: how do you use coding agents for infrastructure as code, pipelines and operations – and how do you provide agents to the whole company in a secure, controlled and cost-efficient way? In agentic coding, platform teams are not just users; they also build the framework everyone else works in.
The dual role of DevOps and platform teams
In most companies, questions about coding agents end up with the platform team sooner or later: Which tools may run in the development environment? What permissions does an agent get in the pipeline? How are credentials protected, and who keeps an eye on token costs? At the same time, DevOps engineers benefit greatly from agents themselves, because much of their work consists of configuration, scripts and recurring adjustments – tasks that are easy to describe and verify.
A workshop for this audience must therefore combine both perspectives: personal productivity with agents and responsibility for a secure agent infrastructure across the company.
Where coding agents speed up DevOps work
The overview below shows typical use cases and the guardrails each one needs:
| Area | Typical agent tasks | Required guardrail |
|---|---|---|
| Infrastructure as code | Create modules, standardise configurations, update versions | Plan or preview only; apply after human approval |
| CI/CD pipelines | Create pipelines, speed them up, analyse build failures | No changes to deployment permissions without review |
| Containers and orchestration | Optimise images, write manifests, set resource limits | Security scans, no privileged containers without justification |
| Monitoring and alerting | Prepare dashboards, alert rules and log queries | Domain review of thresholds |
| Incident analysis | Evaluate logs and metrics, form hypotheses | Read-only access to production systems |
| Scripts and automation | Maintenance scripts, migration helpers, internal tools | Tests and dry runs in non-production environments |
| Documentation | Keep runbooks and operations docs up to date | Cross-check against the actual system state |
Running coding agents securely: the platform perspective
Agents execute commands, read files and access networks. From the platform's point of view, they are a new type of actor that needs clear rules. In the workshop you work out, among other things:
- Isolation: agents run in sandboxes, containers or dedicated development environments, not directly on machines with broad access rights.
- Least privilege: separate identities and tokens for agents with tightly limited rights, distinct from personal accounts.
- Network access: control permitted destinations via allowlists to limit data exfiltration and the download of unknown packages.
- Secrets management: credentials never appear in plain text in an agent's context but come from existing secret stores with short-lived tokens.
- Approval levels: which actions an agent may take on its own and which require human confirmation.
- Traceability: logging of actions, labelling of agent-generated changes, audit readiness.
- Cost control: budgets and limits for token usage per team or project.
We cover these topics in detail in Running coding agents securely: sandboxing, permissions and approvals and The cost of agentic coding: planning and controlling token usage.
Integrating agents into the CI/CD pipeline step by step
A focus of many workshops is using agents in the pipeline, for example for automated reviews, fixing failed builds or dependency updates. A proven approach:
- Pick a use case: start with a clearly scoped, low-risk task such as comments on pull requests or proposals for dependency updates.
- Define permissions: the agent gets only the rights it needs for that task – for example read and comment, but not merge.
- Output as proposals: changes arrive as a pull request or comment, never directly on the main branch.
- Limit cost and runtime: set time limits and token budgets per run.
- Monitor quality: evaluate for a few weeks how many proposals are accepted, adjusted or rejected.
- Expand gradually: grant further tasks and rights only once quality is demonstrably good.
More scenarios are described in Coding agents in the CI/CD pipeline; for mobile apps, Automating app releases with CI/CD covers the basics.
Workshop contents for DevOps and platform engineering
We agree the specific agenda individually. Frequently chosen topics include:
- How coding agents work in the terminal and in the pipeline
- Context files for infrastructure repositories: conventions, forbidden commands, environments
- Infrastructure as code with agents: modules, refactoring, version upgrades
- Troubleshooting builds and deployments with agent support
- Providing agents to development teams: isolation, permissions, secrets, logging
- Connecting internal tools, for example via the Model Context Protocol
- Governance, cost control and measuring the benefit
We place AI developers from real-world practice as trainers who use agents in infrastructure and pipeline environments themselves.
Format and cost
Content, format (in-house or remote) and scope are agreed individually – for example as a compact introduction for the platform team or an in-depth training with exercises in your own environment. Pricing is available on request by e-mail; workshops start from EUR 799 net.
Planning an app? Request a free fixed-price quote in 2 minutesConclusion
DevOps and platform teams benefit twice from agentic coding: agents speed up their own work on infrastructure, pipelines and operations, and these teams create the secure framework in which everyone else works with agents. An agentic coding workshop for DevOps and platform engineering teaches both – practical workflows and robust rules for isolation, permissions, secrets and cost.
Frequently Asked Questions
Should coding agents apply infrastructure changes on their own?
It is advisable for agents to prepare changes and generate a preview, while applying them to production systems only happens after human approval.
How do you protect secrets when using coding agents?
With separate, tightly scoped identities for agents, short-lived tokens from a secret store and the rule that credentials never appear in plain text in prompts or context files.
Is the workshop suitable for teams that want to provide agents centrally?
Yes. Providing agents to development teams with isolation, permissions, logging and cost control is a core part of the workshop.
How much does an agentic coding workshop for DevOps teams cost?
The price depends on content, format and scope and is quoted on request by e-mail. Workshops start from EUR 799 net.