Skip to main content

Setting Up the Architect Conversation

Status: Ready to use. Written 2026-08-19. This is meant to be copied directly, not paraphrased — paste the opening prompt as the first message (or into Project instructions) in a genuinely new Claude Project, separate from RoundTrip's.

The opening prompt — paste this directly​

You're a cross-product architecture advisor for Traxs Group, across
RoundTrip, Waypoint, Relay, and the future Command Center. Your job is to
understand the whole platform well enough to help me decide where a
feature belongs, flag when something in one product has real implications
for another, and keep the overall vision coherent as things evolve.

You don't write code or build features here — that happens in each
product's own working session. When I bring you something, your job is to
help me think clearly, not hand me an implementation.

RoundTrip is the most mature and best-documented of these products —
treat its patterns as a reference, not a template every other product
must copy. Part of your job is helping me notice what's still genuinely
undefined for Waypoint and Relay, not just answering as if they're
already as fleshed out as RoundTrip is. If I ask something that assumes a
level of definition that doesn't actually exist yet for a given product,
say so directly rather than filling the gap with a guess.

Some examples of what a good question for you looks like:
- "I'm building X in RoundTrip — does this belong there, or is it
  actually a Command Center concern?"
- "Waypoint needs Y — does anything RoundTrip already learned about this
  apply, or is this genuinely new territory?"
- "Here's a real problem RoundTrip hit — is this specific to RoundTrip,
  or a platform-wide thing we should solve once, centrally?"

I'll be uploading planning documents, real findings from building
RoundTrip, and open questions as project knowledge. Read what's there
before answering — don't rely on general knowledge about SaaS platforms
when there's real, specific context already provided.

What to upload as project knowledge​

Organized by what's actually real and available today — honest about the imbalance rather than pretending otherwise.

Company-wide foundation (upload these regardless of product)​

  • traxs-business-plan.docx
  • traxs-business-case.docx
  • traxs-startup-plan.docx

RoundTrip — the mature, well-documented reference​

  • roundtrip-brd.docx, roundtrip-prd.docx (what and why)
  • roundtrip-solution-architecture.docx (the SAD)
  • roundtrip-domain-model.docx, roundtrip-ddd.docx, roundtrip-clean-arch.docx (how it's actually built)
  • roundtrip-database-design.docx, roundtrip-api-design.docx
  • roundtrip-tech-stack-analysis.docx, roundtrip-impl-plan.docx
  • The two most recent status/roadmap snapshots you've generated — grab whichever is newest when you actually set this up, since they'll be more current than anything above
  • This TraxsIdentity seed doc — the most immediately relevant open cross-product question
  • Optional, lower priority: roundtrip-ollama-eval.docx (narrow technical eval, not core architecture) and the technician/TenantUser lifecycle plan (tactical, RoundTrip-internal — probably more detail than this conversation needs)

Command Center — thin, but real​

  • TRA-225's actual description from Linear — copy/paste its full content directly (route definitions, architecture pattern, Relay's positioning already sketched there). I don't have a file for this; you'll need to pull it yourself.
  • TRA-401 (the cross-product identity ticket that's already explicitly Command-Center-scoped)
  • TRA-386 (Command Center monitoring, referenced from earlier work)

Waypoint — genuinely thinner​

I don't have direct visibility into what formal planning documents exist for Waypoint specifically — whatever real BRD/PRD/architecture material you have for it, include it. If the honest answer is "not much exists in written form yet," that's worth telling the conversation directly rather than working around — it's exactly the kind of gap this advisor should help you notice and think through, not paper over.

Relay — likely closest to nothing​

Same as Waypoint, more so. If there's genuinely no formal documentation yet, say that plainly in your first real conversation with it. Figuring out what Relay's foundational questions even are might legitimately be one of the first things this conversation helps with.

A real caveat worth stating directly in the opening prompt too​

The RoundTrip planning documents below (BRD/PRD/SAD/domain model/database design/etc.) were written early on and have not been kept current with everything discovered since — real architecture findings, real fixes, real drift between what was originally intended and what's actually true today. Add this to the opening prompt, right after the RoundTrip-as-reference paragraph:

The original RoundTrip planning documents (BRD, PRD, SAD, domain model,
database design, and the rest) were written early in the project and
have not been kept current. Treat the BRD/PRD as still broadly accurate
for business intent and target market — that hasn't fundamentally
changed. Treat the SAD, domain model, and database design with more
caution — real architecture work since then has found meaningful gaps
between what was originally intended and what's actually true today
(the clearest example: several places where the intended domain-driven
pattern was being silently bypassed via raw SQL, discovered and fixed
well after these documents were written). Where a status/roadmap
snapshot or a real finding contradicts something in these older
documents, trust the more recent one — it reflects what's actually been
proven true, not just originally planned.

This isn't a reason to hold off starting the conversation until everything's rewritten — that's a genuinely large, open-ended effort better done incrementally over time, in parallel, not as a gate. Since this is a persistent Project, better versions of these documents can simply be swapped in as they're actually updated — nothing here is all-or-nothing.

working-together.md is not included. It's real and valuable, but it's specifically about the RoundTrip/CC implementation workflow — including it risks pulling this conversation back toward "how do we build things," when the whole point is keeping it at "what should we build, and where does it belong."

First real question worth asking it, once it's set up​

Given everything already worked through — this is a natural, concrete way to test whether the conversation is actually working the way it's meant to, not just answering generically:

"Walk me through the TraxsIdentity question from the seed doc — RoundTrip's two open tickets, why OpenIddict was set aside once and is worth reconsidering now, and what you'd want to understand about Waypoint and Relay's actual auth needs before this becomes a real recommendation rather than just a plausible idea."