A Proven Agentic Coding Workshop Agenda (Half-Day, One-Day, Multi-Day)

The agendas below are templates you can adapt for an internal workshop or use to compare provider offers. They assume a group of developers using one approved coding agent such as Claude Code, OpenAI Codex, Cursor or GitHub Copilot, working on their own repository or a realistic sample. Times are guidelines; the ratio of hands-on to presentation time matters more than the exact minutes.

Design principles behind every agenda

• At least 60 percent hands-on. Participants learn agent workflows by running them, not by watching demos.
• Real tasks. Labs use tickets from the team's backlog or a sample repository in the same stack.
• Every block produces an artifact. An instruction file, a pull request, a checklist, a configuration file.
• Verification is never optional. Every lab ends with tests run and a diff reviewed.
• Policy in context. Security and data rules are covered where they apply, not in a separate lecture at the end.

Half-day agenda (about 4 hours)

Use this as an introduction for teams new to agents, or as a taster before deciding on a longer program.

What to cut if time is short: shorten the theory block, never the lab.

One-day agenda (about 7 hours)

This is the format most teams should start with. It covers the full loop from specification to merged pull request.

Multi-day program (2 to 3 days plus follow-up)

For teams that want to go beyond interactive use toward automation and background agents. Day 1 follows the one-day agenda. Further days add:

Day 2: Extending the agent

1. Reusable commands and subagents for recurring workflows such as migrations, release notes or code review.
2. Hooks and deterministic guardrails: formatters after edits, blocked paths, command restrictions.
3. MCP integrations: connecting the issue tracker, documentation or a read-only database, with a discussion of which servers are trusted.
4. Refactoring lab: characterization tests first, then an agent-driven refactor of a messy module.

Day 3: Background agents and CI

1. Environment setup so cloud or headless agents can install dependencies and run tests.
2. Delegating tickets asynchronously and limiting work in progress so reviews keep up.
3. Headless runs in CI for narrow chores, with tight permissions.
4. Metrics: deciding what to track over the next weeks (see measuring the ROI of agentic coding training).

Follow-up session (2 to 4 weeks later, 2 to 3 hours)

Participants bring problems from daily use. Typical topics: instruction files that have grown too long, review bottlenecks, tasks where agents consistently fail, and policy friction. This session often delivers as much value as the main workshop, because it addresses real rather than anticipated issues.

Adapting the agenda to your context

• Mixed tools in one team: add 30 minutes on keeping instruction files and rules consistent across agents. AGENTS.md is read by several tools, which helps.
• Legacy code with few tests: replace Lab 1 with "let the agent write characterization tests for a module."
• Mobile teams: include build and signing commands in the instruction file and plan for slower build cycles in labs.
• Remote delivery: split the one-day agenda into two half-days; see on-site vs. remote AI coding workshops.
• Enterprise setting: start with the approved toolchain and managed settings, and add a short session with information security.

Preparing participants beforehand

A short pre-work email saves the first hour. Ask participants to install the agent and confirm their account works, check out the repository and run the test suite once, and bring one small ticket they would like to delegate. Organizers should confirm that the build works on participants' machines or in the chosen cloud environment, and that everyone knows which data and repositories are allowed. If the build takes longer than a few minutes, plan labs around smaller modules or prepare a pre-built environment.

Checklist for comparing provider agendas

• Is at least half the time hands-on?
• Do labs use your stack and, ideally, your code?
• Is there explicit time for reviewing agent output?
• Are security and data rules covered in context?
• Does the team leave with committed artifacts?
• Is a follow-up session included or available?

neooutput builds its agentic coding workshops on this kind of build-as-you-learn structure and adapts agendas to the team's tools and stack. For the content side of what each block should teach, see what a good agentic coding workshop for teams covers. To discuss an agenda for your team, book a free intro call.