MWP Sprint. From zero to production in 4 to 6 weeks.
End-to-end implementation of the Model Workspace Protocol in your operation. You end with a system running in production, your team trained, and a complete handoff. Scope documented before we start, no loose hours.
What it is
In 4 to 6 weeks you go from zero to a running system: a business agents.md, MWP folder architecture, MCPs connected to your stack (Gmail, Calendar, CRM, Notion, Supabase or whatever you use), skills per department, RAG over your documents, n8n workflows, and a dashboard with usage metrics. We don't sell a deck or hours: the Monday after go-live the system works and your team operates it.
The method is the Model Workspace Protocol (Van Clief and McDermott, arXiv 2026): folder structure as agent architecture. Everything we build is interpretable, debuggable, versioned in git, and portable. The same thing that makes the system shippable makes it yours.
Who it's for
- LATAM companies of 20 to 80 people with 1 to 3 internal technical staff
- Teams with quantifiable pain: they can't scale without hiring, or broken processes with a real cost in hours or money
- Operations that validated the map with the Audit, or arrive with a clear process to automate
- A CTO or technical lead as project champion
What you get
- 01Business agents.md: context, voice, and formalized processes
- 02MWP folder architecture adapted to your real workflow
- 03MCPs connected to your stack: Gmail, Calendar, CRM, Notion, Supabase or whatever you use
- 04Skills per department: executable SOPs, not dead PDFs
- 05RAG over your business documents (pgvector on Supabase)
- 06Self-hosted n8n workflows for core processes
- 07Dashboard with usage and return metrics
- 082 training sessions with your team and an operations runbook
How we work
- Week 1
Immersion
Workflow mapping with your team, context audit over your real operation, MWP architecture design.
- Week 2
Quick win
One functional workflow in production, visible to the team. Builds momentum and validates the approach.
- Weeks 3–4
Core build
MCPs connected to the stack, skills per department, RAG populated, dashboard with metrics.
- Week 5
QA and training
End-to-end testing of the full system and the first training session with your team.
- Week 6
Go-live
Complete handoff, second training session, operations runbook delivered.
What it does NOT include
- It does not include evolutionary maintenance after handoff: that is the Operations Retainer.
- It does not build on unmapped processes: if the map isn't clear, we start with the 60/30/10 Audit.
- It does not deliver tools your team can't operate: if something needs us to keep working, it was built wrong.
Specific questions
What does guaranteed go-live mean?
That if anything in the agreed scope doesn't end up running in production, we refund the proportional part. It's documented in the agreement before we start.
Who owns the code, the agents.md, and the MCPs?
You do. Everything stays in your repo, your infrastructure, your accounts. MWP is plain text versioned in git: you can hire someone else tomorrow and nobody depends on us to understand the system.
Do we need a technical team?
Ideally 1 to 3 internal technical people to operate the system after handoff. Without them the Sprint still works, but the Retainer becomes a condition for success at 12 months.
What does the scope depend on?
On the number of MCPs to connect, stack integrations, and departments entering the system. We define it together in the initial conversation and it's fixed in the agreement, no loose hours, no surprises.