Fast alone, slow together
One engineer with an agent can finish a task before lunch. Put 6 engineers on the same codebase, each with their own agent habits, and the seams start to show. Specs live in chat windows, the architecture drifts a little with every pull request, reviewers drown in volume, and by Friday nobody's quite sure what's actually done.
Nobody did anything wrong here. The tools showed up before the process did, and almost every team using agents is living somewhere in that gap right now.
What your team walks away with
Specs an agent can build against
Acceptance criteria written and approved before the first line of code.
Work split for parallel hands
People and agents on separate tasks and branches, without stepping on each other.
A real definition of done
Your own build, test, and eval commands as checks that block done.
Review that keeps up
Review at the grain that fits (task, feature, or epic) so volume doesn't bury the team.
A person on every merge
Agents carry the work to the line. A human gets it across.
Status without a meeting
One board and one report everyone reads, from the work itself.
One real feature, plan to merge, in your repo
We don't teach on a toy app. Your team brings a real feature from its backlog, and we take it through the whole passage together.
- Agent Decompose
- You Approve the shape
- Agent Specify
- You Approve the plan
- Agent Build and pass checks
- Agent + you Review
- You Merge
The highlighted steps are gates. A person decides at every one of them.
Three ways to start
Leadership Briefing
Half day · up to 12 leaders · remote or onsite $7.5K Credited toward a Team WorkshopFor CTOs, VPs of Engineering, and PE operating partners. What changes in software delivery when agents write the code, what to measure, and where teams tend to get hurt. You leave with a read on where your team stands today.
Team Workshop
2 days onsite or 4 half days remote · 6 to 16 people · 2 instructors $25K to $35KThe core program. Your team takes one real feature from plan to merge in your own repo, with your own agents and your own checks. You keep the plan, the wired-up checks, the merged feature, and a 30-day plan for what's next.
The Passage
Workshop + 6 weeks of coaching $60K to $90KFor teams that want the habits to stick. After the workshop, a Conductor joins your weekly planning and review, reviews specs and pull requests between sessions, and closes with a readout on how the team's delivery changed.
Not ready for a team workshop? Start with Tiller on your own.
We teach with Tiller. You don't have to keep it.
We teach with Tiller because we built it, it's free and open source, and it makes the method something you can see. But the habits are the point. Specs before code, checks that block done, and a person's hand on every merge work with whatever your team runs next.
Bring the agent you already use. Tiller works with Claude Code, Codex, and Pi today, and with a plain shell.
Read the Tiller docsTaught by people who ship this way
Every session is led by Made In Tandem's senior engineers (Conductors), the same people who built Tiller and use it on client work.
JC Grubbs
Founder & CEO · ConductorJC has been writing software for more than 20 years and still works as a hands-on Conductor on client engagements. That work runs from product scoping and architecture to putting coding agents into real delivery: writing the specs they build against, wiring up the checks that block done, and reviewing what they ship. JC built Tiller to make that way of working repeatable for a whole team.
Allen Dayag
ConductorAllen has spent more than 10 years building web and mobile apps. As a Conductor, he owns client engagements end to end: shaping the product with the client, setting the architecture, and keeping people and coding agents on one plan through review and merge.
Where it fits, and where it doesn't
A good fit
- A team of 3 or more engineers already using coding agents, unevenly.
- A product you can describe before an agent builds it.
- Someone outside the team needs to see where things stand.
Probably not a fit
- Solo developers (Tiller's quickstart is free and a better start).
- Research spikes and throwaway prototypes.
- A team looking for a tour of one tool's features.
Questions teams ask before they book
Do we have to use Tiller after the training?
No. We teach with it because it makes the method visible, and it's free and open source if you want to keep it. The habits work without it.
Which coding agents does this work with?
Bring the one you already use. Tiller works with Claude Code, Codex, and Pi today, and with a plain shell.
Remote or onsite?
Either. Onsite runs as 2 full days. Remote runs as 4 half days, which most teams find easier to absorb.
How many people can attend?
The Team Workshop works best with 6 to 16 people and 2 instructors. Larger teams can run in groups.
What do we need to prepare?
One real feature from your backlog, access to your repo with the agents and tools your policies already allow, and the people who own that feature's requirements and review. We do a short prep call and a look at your repo before day 1.
Is our code safe?
Your code stays in your environment, and we work under your approved tools and policies. Our AI governance practices are at /ai-governance/.
What happens after the workshop?
You keep everything we built together: the plan in your repo, the wired-up checks, the merged feature, and a 30-day plan. Teams that want ongoing help choose The Passage.
How is this different from Claude Code or Copilot training?
Tool training teaches one person to drive one agent. This teaches a team to plan, build, check, and merge together without drift, whatever agent each person uses.
Let's talk about your team
Tell us how your team works today and which format you're leaning toward. A Conductor reads every note and replies with a straight answer on fit, timing, and next steps.