We make software that feels inevitable.

Monospace is a design and engineering studio. We design and build product, platform, and AI-native software — web platforms, native macOS and iOS apps, design systems, internal tools, and 0→1 product work. Design and engineering are done by the same hands, so there is no handoff between a design file and the people who ship it.

The base rate is €2,500 per week, and longer commitments are discounted against it. A sprint is two to four weeks at €5,000 – €10,000 as a fixed fee. A build runs one to six months at €10,000 – €50,000 — the ceiling is well under the weekly rate times six months, because larger projects carry a volume discount. An embed is quoted per engagement rather than banded: the price turns on days per week and how long the commitment runs, with a three-month minimum. Committed capacity is discounted against the weekly rate, so it works out cheaper per week than a sprint. Every engagement is scoped and priced before work starts — there are no hourly surprises.

A sprint is two to four weeks of focused execution: brief in, deliverable out. Best for prototypes, audits, and sharply scoped builds. A build is one to six months of end-to-end product creation, from research and design through engineering and launch. An embed is continuous — we join your team and ship alongside it, two days a week up to full time, and it is the cheapest of the three per week because the capacity is committed.

A sprint delivers a working artifact in two to four weeks. Every stage of the process ends in something tangible rather than a deck: discovery ends in a scope, design ends in a prototype in code, engineering ends in production software. Live software beats perfect plans, so we deploy early and iterate against real usage.

Not chatbots. It means orchestration, billing, and observability for production agentic systems — multi-agent routing with confidence tags and explicit handoff markers, human-in-the-loop execution where AI proposes as structured previews and a hardened executor applies with validation and audit trails, and Zod schemas around every model call so output cannot drift past its contract. Shipped in Resolo, Saldo, and Checkmarked.

OpenAI, Anthropic, and Google under one orchestration layer, with token-to-credit ledgers per organization and attribution across every call. Model choice is a routing decision, not a lock-in — the orchestration layer sits above the provider so a model can be swapped without touching product code.

TypeScript and Next.js on the web, Tauri with Rust for native desktop, Swift and SwiftUI where the platform demands it, and Postgres with Redis and pgvector behind them. The stack is chosen per project, but these are the defaults because they are what we have shipped and can support.

Both, by the same people. Interfaces are prototyped in code rather than mocked up, because real interfaces beat static screens and the prototype becomes the product. Brand and design systems — tokens, components — are engineered to scale rather than delivered as a static file.

Five stages, each ending in an artifact. Discover: stakeholder interviews, technical audits, and scope. Design: sketch, prototype, refine, in code where possible. Engineer: production-grade systems from day one, type-safe and observable. Ship: deploy early and often. Measure: iterate against real usage, with numbers rather than opinions.

Monospace is remote and available worldwide. Work runs asynchronously by default, with calls booked directly against a live calendar rather than negotiated over email. Contact is hello@bymonospace.co.

Rust and Tauri for native desktop, because a Mac app should launch instantly and work offline. TypeScript and Next.js on the web, because the type system is the cheapest test suite available. Postgres with Redis and pgvector behind both. Every choice is one we have shipped and can support — Ampersand, Pastry, and Cali are Tauri and Rust; Saldo, Resolo, and Checkmarked are TypeScript and Postgres.

Both, and the products come first. Ampersand, Ola, Pastry, Cali, and All Things Claude are studio-owned software shipped to real users. Building our own products is how the studio finds out what an engineering decision actually costs over a year rather than over a sprint.

The model never holds the write path. AI proposes as a structured preview, and a separate hardened executor validates it against a schema, checks permissions and tenancy, and applies it inside a transaction — with an audit trail recording what was proposed, what was applied, by whom, and which model and prompt version produced it.

It depends on where the data belongs. Clipboard history, documents, and calculations stay on the device — Pastry uses local SQLite, Ampersand writes plain Markdown to a folder you choose, Cali writes portable .cal files. Multi-tenant platforms are cloud systems because they have to be. The test is whether the data would be safer or more useful somewhere else.