Thinking in Agents

Practical agent workflows for engineering teams

Your team learns the mental models that make agent workflows work — regardless of which tool comes next. Hands-on with current tools, but what stays is the system behind them.

5–20 TeilnehmerMärz April — Nächster freier Termin: Mai

Who it's for

Engineers who want to move beyond autocomplete into real agent workflows — at their own codebase.

Tech leads who need a framework for when to use agents, when to go manual, and when to stop.

CTOs and decision makers who need clarity on what agents can actually do today, where the ROI is, and where the risk lies.

Why teams book

Mental models that outlast tools. Agent loops, context management, delegation patterns — your team understands the system behind the results. That stays when tools change.

Hands-on at your own code. Everyone builds workflows on real project context, not slide exercises.

Production-ready from day one. Tests, linting, CI and clear guardrails — so it doesn't stay an experiment after Monday.

One day, five phases

  1. 01
    The Shift — Vibe coding vs. agentic engineering
    Spectrum: autocomplete to orchestrator. Taste over syntax. Live demo: same task, different approach.
  2. 02
    How Agents Work — LLMs, tool calls, agent loop
    Context windows & failure modes. Agent loop: think, act, observe, repeat. Token economics & model tiers.
  3. 03
    Hands-on — First workflows at your own code
    Harness setup & model selection. Project instructions via CLAUDE.md. Trust boundaries & intervention points.
  4. 04
    Guardrails — Tests, CI and back pressure
    Tests as first-line defense. CI as organizational back pressure. The trust equation for agent autonomy.
  5. 05
    Scale — Orchestration, enterprise & adoption
    Parallel agents & task splitting. MCP servers. Data policies & agentic readiness scoring.

Half-day: phases 1-3. Full-day: all 5 phases. Content customizable on request. Full curriculum →

What your team takes away

Working workflows. Everyone leaves with agent workflows that run on their own repo on Monday.

Project setup. CLAUDE.md, custom commands, and harness configuration — tailored to your stack.

Decision framework. When agent, when manual, when stop. No guesswork, no overcommitment.

Investment

Formats Half-day or full-day
Offer on request
Group size 4-20 people
Region DACH (in-person), remote on request
Language German or English

In-house in DACH, including preparation and follow-up. Travel costs as incurred.

Get in touch

Pick a format, pick a date. I'll tailor the content to your stack and goals.

← back to workshops
12 posts tools lexicon follows rss © 2026 siever.ing