LATEST NOTE I built a night agent for myself

We engineer AI-native products that ship.

Crewtives is a small studio building production-grade software where agents are first-class citizens — strategy, design, engineering, and the operator's CLI that ships them.

Currently Available
Next slot June 2026
Remote Worldwide

01 / Selected work

The kind of things we ship.

02 / Behind the studio

Migue San Martin — Founder & engineer
NAME Migue San Martin ROLE Founder & engineer HANDLE @crewtives

A note from the founder.

I started Crewtives because I couldn't find a studio that actually built the whole thing. Most agencies hand you Figma files and call it a day. Most contractors fix the bug in front of them and ignore the system around it.

I wanted to work the other way: own the strategy, the design, the engineering, and the operator's CLI that ships them. Treat the product as one thing, not five contracts stapled together.

If you're building something where humans and agents both need to drive the same product — research tools, AI-native dashboards, anything that's complicated enough to need an MCP server next to the React app — this is what we do. Crewtives is currently me and a small group of trusted collaborators. We take a couple of projects at a time. Let's talk.

migue san martin @crewtives

03 / How we work together

Three ways to start a project.

Three shapes that cover most of what we take on. All engagements are fixed-scope — no time-and-materials surprises. We agree the price in the kickoff, after we've scoped the work together.

Discovery sprint

1 week

We pressure-test the idea, write the KICKOFF doc, scope the build, and ship one working prototype.

  • Best when…
  • Validating an AI feature idea
  • De-risking a new product line
  • Replacing a vendor / tool
Discuss this

Embedded retainer

rolling, 1 mo min.

We embed into your team. Shared Slack, PR reviews, weekly pairing. You get senior engineering on tap.

  • Best when…
  • Existing team needs senior AI eng
  • Scaling product after launch
  • Stuck on an evals/agent problem
Discuss this

04 / Process

We work like a small product team.

Every project leaves your team with the docs, the CLI, the agent prompts, and the evals — not just the code.

  1. 01

    Kickoff

    We write a KICKOFF doc together — what we're building, what's locked, what trade-offs we agree to. The doc is the contract, not a deck.

    KICKOFF.md
  2. 02

    Build & ship

    Weekly milestones, weekly demos. AI personas review every PR before humans see it. You watch the build log in real time.

    main · prod
  3. 03

    Measure

    Each release ships with the metric it's supposed to move. We instrument first, ship second — so we can tell if it worked.

    pulse-reports/

05 / Stack

The eight tools we reach for first.

06 / FAQ

Questions clients actually ask.

01 How quickly can you start?

We take a small number of projects per quarter. Discovery sprints can usually start within 1–2 weeks; build engagements within 3–4 weeks. If we're booked, we tell you straight away and recommend someone we trust.

02 How much does a typical project cost?

We agree pricing in the kickoff, after we've scoped the work together. Fixed scope — no time-and-materials surprises. If scope changes mid-flight, we write a new SOW so you always know what you're paying for.

03 What stack do you work in?

TypeScript everywhere, Astro for landings, Next.js for app workbenches, Rails for heavy backends, Postgres for state, Cloudflare for edge. Model layer: Claude, OpenAI, plus open-source models via Ollama/vLLM when self-hosted is required. Agent layer: MCP servers, LangGraph, custom loops.

04 Do I own the code?

Yes. All code, models, prompts, evals, and infra config are yours from day one. We don't keep license clauses. We do reuse our own internal patterns across projects — but anything specific to your product is your IP.

05 Are you really a studio, or just one person?

Crewtives is currently led by one founder with a small group of trusted collaborators we bring in per project (design, motion, backend specialists). For most engagements you'll work directly with the founder. For larger builds we scale up transparently.

Let's talk

Tell us what you're building.

We take on a small number of projects per year — usually 0-to-1 AI products, agent-native rewrites, or design-system rescues. If that's near what you need, let's talk.

Currently taking projects · next slot: June 2026
  1. What happens next
  2. 01

    Reply within 24h

    We acknowledge every email within one working day, even if it's a 'not the right fit'.

  3. 02

    30-min intro call

    We discuss what you're building, what success looks like, and whether we're the right team for it.

  4. 03

    Scoped proposal

    Within a week of the call, we send a fixed-scope proposal with timeline, deliverables, and price.

08 / Notes

Build log & field notes.

Short essays from inside the studio. Each is something we learned in the field — no listicles, no SEO, no LinkedIn cadence.

All notes