I started Typiva because choosing a type direction meant jumping between disconnected tools, and none took a designer from a brief to a usable brand system.
I designed and built a typography product that turns one brief into a direction, pairings, alternatives, and a brand kit. Since its 2025 launch, it has reached 1,000+ visitors and shipped through two public UI refreshes.
Most font tools answer one question. Font identifiers tell you what's on a page. Font generators show you what something looks like set in 12 faces. Font finders sort a database by a single attribute. None of them walk you from a brief to a confident decision, and none of them care that you're using the result across a marketing site, a product, and a deck that all need to look like the same brand.
What I wanted was a different shape entirely: type a brief, get a direction. That single shift — brief in, direction out — meant the tool had to think about pairings, alternatives, and brand consistency as part of the answer, not as separate features the user had to stitch together.
I designed and built Typiva end-to-end as a solo, self-funded product. MagicPath carried the early exploration and UI direction. Lovable got the first working version of the product out fast — the brief-to-pairings pipeline shipped in Lovable before I rewrote the parts that needed to be more careful. Cursor and Claude Code took the codebase from prototype to production. Lovart did the logo and brand work alongside the product.
Two public UI refreshes shipped in twelve months, each one faster than the last. The first tightened the visual system and clarified the brief input. The second rebuilt the landing and product UI to match what I'd learned about how designers actually use the tool — most people come in with a rough direction, not a finished brief, and the input shape had to match that.
Once the core product was stable, I built typiva-mcp — a Model Context Protocol server that gives AI agents the same typography primitives Typiva uses. The reason was selfish: I was already using AI agents to draft UI, and I was tired of them picking generic type pairings because they had no other option.
The MCP exposes the design tokens (colour, radius, shadow) and the font pairings, alternatives, and brand-kit generation flows as tools an agent can call. Now any agent working in the same codebase can pull from the same typographic vocabulary the product is built on — not the default whatever-the-model-thinks-is-nice. That's the deal: the agent's taste matches the product's taste because they're reading from the same file.
Live at typiva.com — brief in, pairings out, brand kit generated, MCP available for any agent that wants it.
Framed the product for designers and builders who already have context — not for type specialists. The brief-in, direction-out shape was the actual product, not the database. People don't want a list, they want a confident answer they can defend.
Search, pair, and generate a brand kit from a single prompt. I deliberately stayed inside the free-font universe — no paid-fonts source, no licensing tier, no upgrade path — so designers could ship a complete brand without licence anxiety on day one, and so the catalogue stayed within indie-product scope instead of bloat.
AI-native doesn't mean AI-generates. It means the human stays in the loop and the tools compound the work. The first AI pairings were confidently wrong in subtle ways — I rewrote every one before it shipped.
People don't want a list — they want a direction. The brief-to-font pipeline was the actual product, not the database. The database was the raw material; the pipeline was the value.
Indie products move at indie speed. No roadmap committee, no PMO. The constraint forces a faster feedback loop: ship a refresh, watch what people actually use, ship the next 10% better each pass.
If I were starting over, I'd build the MCP from day one. The agent-first version of Typiva would have made the design-token discipline part of the product shape from the first commit, not an after-thought I had to retrofit.
