Custom builds
Bespoke AI software, built on Claude.
Internal copilots, ops dashboards, customer-facing AI products and web apps. Designed for the workflows your business actually runs on - and built by an operator team that ships its own software in production every week.
Built by an operator who ships
What we build
Three flavours of build. Same operator team behind all of them.
Internal AI tools
Bespoke copilots, dashboards and ops tools your team uses every day. Built on Claude with the access controls, audit logging and integrations your business actually needs.
Examples
- Sales copilots over your CRM
- Ops dashboards with AI summaries
- Internal knowledge assistants (RAG)
- Approval and triage agents
Customer-facing AI products
AI shipped inside your product or website. From chat assistants to fully agentic workflows - prompt-cached, tool-using, production-grade.
Examples
- AI-powered onboarding flows
- Support assistants with tool use
- Voice agents on Twilio / phone
- Embedded AI features in web apps
Web apps and SaaS
Modern web apps and small SaaS products. TypeScript + React + Cloudflare or Supabase stack, with Claude built in from day one - not bolted on later.
Examples
- Internal SaaS for a single business
- Marketing sites with AI features
- Multi-tenant tools for partners
- Replatforming legacy internal tools
The stack
Modern, pragmatic, fast to ship.
Language models
Claude (Opus, Sonnet, Haiku) by default. Prompt caching, model routing and clean migrations between versions baked in.
Frontend & backend
TypeScript, React, TanStack, Tailwind, Cloudflare Workers, Supabase. Pragmatic, modern, fast to ship.
Agents & tools
Anthropic Agent SDK, MCP servers, custom tools wired to your CRM, billing, scheduling and internal APIs.
Data & RAG
RAG over your docs, tickets and product data. Vector + lexical, with the citation discipline your legal team will sign off on.
How it works
Discovery to v1 in 4-12 weeks.
01
Free 30-min discovery call
Walk us through the workflow you want to build for. We'll confirm fit and propose the smallest possible v1 that still delivers value.
02
Scope and quote
Fixed-scope quote with a milestone-based timeline. Most v1 builds ship in 4-12 weeks. No multi-month discovery phases.
03
Build with weekly demos
Working software in weekly demos from week 1. You see progress in production, not Figma mockups - and we course-correct fast.
04
Ship, train and hand over
We ship to production, train your team and document everything. You can run it, extend it, or have us stay on for ongoing improvements.
FAQ
Common questions.
What's the typical price range?+
Most v1 builds land between $25k-$120k AUD depending on scope. We don't publish a public price list because every build is different - we'd rather give you an honest fixed quote after a 30-minute discovery call than over-promise here.
How long does a build take?+
Most v1s ship in 4-12 weeks. We deliberately scope to the smallest version that proves value, then iterate from there. Multi-quarter builds are possible but rare - usually we'd rather ship a v1 fast and grow it.
Who owns the code?+
You do. We hand over the repo, the deployment, the documentation and the operational runbooks. No lock-in, no vendor games.
Do you build on something other than Claude?+
We default to Claude because we know it deeply and it's our partner application focus. If your build genuinely needs another model (e.g. a self-hosted open-weight model for a specific compliance constraint), we'll tell you up front and either help you scope it or refer you on.
What about ongoing support and maintenance?+
Optional. Some clients take the build and run it themselves. Others put us on a monthly retainer for ongoing development, model migrations, and small feature work. Both are fine - we don't push lock-in contracts.
Can you work with our existing dev team?+
Yes. We'll work alongside your engineers, do code reviews against your conventions and hand over cleanly. We can also install Claude Code into their workflow at the same time so the build skills compound.
Related reading
Operator notes on building custom AI software.
When custom builds pay back, what to wire into the SDK, and the agentic patterns we ship into production.
Custom Builds · 8 min read
Voice agents for SMBs: when they pay back (and when they're a vanity project)
AI voice agents on the phone are getting genuinely good. Here's the honest read on where they save real money for an Australian SMB - and where they're a six-figure project that should be a chatbot instead.
Read article →
Explainers · 7 min read
OpenRouter and multi-model routing: why no serious AI workflow runs through one vendor
Why piping AI workflows through a router beats single-vendor lock-in: cost arbitrage, model fallback, vendor concentration risk in diligence, and what to actually configure.
Read article →
Explainers · 9 min read
Agentic AI for operations: what an 'agent' actually is, and where it works
An agent is a loop, a set of tools, and a memory — not magic. Where agents reliably remove cost in back-office ops, where they break, and how to scope a first one that survives diligence.
Read article →
Book a free 30-min discovery call
Tell us what you want to build.
Bring the workflow, the rough idea, or just the problem. We’ll come back with an honest read on fit and a v1 scope you could actually ship.