We build AI agents for
small and medium business.
Agents that take on the repetitive work — built into the tools your team already uses, without the technical headache.
What we build.
From agents to production.
Four focused capability areas where senior AI engineers move faster than any in-house build.
AI Platform
Build and deploy custom AI agents through a unified platform — a simple dashboard and one API endpoint for every app.
TryVoice
AI voice agents that answer your business phone on an Australian number and take bookings — powered by Lightspire.
TryRAG Solutions
Connect your documents and data to AI. Get accurate, context-aware answers grounded in your own information.
TryAI-Accelerated Development
Ship software 80% faster with AI tools that deeply understand your codebase, architecture, and coding patterns.
TryAgents build and test.
People release.
Infonex Open Factory is our open-source software factory for AI coding agents. Agents take each ticket to a running test environment with no human in the loop, and nothing reaches production until a person approves it.
Agents only · no human in the loop
Human release
Zone one, agents only, with no human in the loop. Step 1, intake: a ticket, an epic or one line of instructions. Step 2, analyse: scope, risk and size. Step 3, spec: an OpenSpec change plus hidden scenarios. Step 4, build: several developer agents work in parallel. Steps 5 and 6 run in parallel on the same commit: verify runs the app against the hidden scenarios, and review checks the change with a fresh context and no developer notes. If either fails, a fix round sends the findings back to build, up to two times. Step 7, merge to main, one change at a time once CI is green. Step 8, the test environment is deployed from main automatically.
Then a release gate. Zone two, human release. Step 9, a person reviews the test environment and approves. Step 10, production: your team releases it.
Hidden scenarios
The spec writer also writes 3 to 8 holdout scenarios that the developers never see. Only the verifier runs them, against the running app.
Independent review
The verifier and the reviewer start fresh, without the developers' notes. The reviewer checks spec coverage, bugs, security and gate tampering: skipped tests, loosened lint, edited CI.
Merges, never deploys
Changes merge to main one at a time, after CI passes. The factory never deploys, tags or releases to production. That stays with your team.
Works with your coding agent
Built on Claude Code. Codex, Gemini CLI, GitHub Copilot and Cursor read the same Agent Skills format, so they are supported, but not yet tested.
Our research.
Building AI in production.
Plain-English field notes on the techniques behind our work — the formulas and trade-offs that actually hold up with real data.
RAG: getting the right context, every time
Most wrong answers aren't a model problem — they're a retrieval problem. Here's the formula we landed on for grounding AI in your own data.
Read researchAI agents: when to let the model take actions
Agents are powerful and easy to get wrong. The formula that worked for us: constrain the tools, verify every step, and keep a human in the loop where it counts.
Read researchVision: turning documents and images into structured data
OCR alone is brittle. The formula that worked: pair a vision model with a strict schema and a confidence check, so messy scans become clean, trustworthy data.
Read researchEvaluation: how we know an AI feature actually works
You can't improve what you don't measure. The formula that changed everything for us: a golden dataset and an LLM-as-judge, run on every change.
Read researchThree steps to working software.
No theatre, no decks.
Discover
We map your workflows, codebase, and existing systems — finding the highest-impact place to add AI.
Ship
We build alongside your team using AI-accelerated workflows. Real code, in your repo, on your stack.
Scale
We harden, instrument, and hand off — or stay on as your long-term AI engineering partner.
Specialist services.
From senior AI engineers.
Deep expertise across every layer of the modern AI stack — from strategy and integration through to production agents and training.
AI-Accelerated Development
Build software 80% faster with AI tools that understand your codebase, architecture, and coding patterns.
Learn moreAI Agents Development
Custom AI agents that reason, decide, and act — automating complex workflows that previously required human judgment.
Learn moreLLM Integration
Integrate large language models into your existing apps. Add chat, summarization, classification, and AI-driven features.
Learn moreRAG Solutions
Build an internal knowledge base your AI can talk to. Accurate, context-aware answers grounded in your own data.
Learn moreMCP Development
Expose your internal APIs to AI using the Model Context Protocol. Let AI models securely access your tools and data.
Learn moreAI Consulting & Strategy
We assess your business, identify high-impact AI opportunities, and build a clear roadmap from strategy to execution.
Learn moreCustom Software Development
Full-stack development powered by AI-accelerated workflows. We build web apps, mobile apps, APIs, and enterprise systems.
Learn moreSpec-Driven Development
Turn product ideas into structured specs that AI can build. From planning to code — faster, clearer, and more predictable.
Learn moreAI Training & Enablement
Hands-on workshops, structured courses, and ongoing support to make your entire organisation AI-ready.
Learn moreWe've shipped across
the most regulated industries.
Trusted by teams
shipping complex systems.
Delivered critical system integrations with excellent communication and technical precision.
Deep .NET expertise and seamless team integration on our high-impact project.
Outstanding professionalism and technical skill during our infrastructure project.
Latest from the team.
Notes from the build.
- · AI Engineering · 3 min read
PageTruth Is Now Open Source: Answers That Open the Exact Page
PageTruth is a team knowledge base that answers only from your documents and cites the exact page. We've open-sourced it under MIT. Here's how it works, what our evals found, and how to run your own.
- · AI Engineering · 3 min read
The Wiki That Never Goes Stale: Let Your Coding Agent Own the Docs
Project knowledge is scattered and every wiki drifts. Our proposal: a small markdown wiki inside the repo that your coding agent reads before each change and updates after it — with a sample repo to clone.
- · AI Engineering · 3 min read
8 Ways to Cut Your Token Usage in Claude Code
Claude Code bills by the token, and almost all of that cost is context you keep re-sending. Here are eight habits that shrink the bill without slowing you down.
Ready to build something great? Let's scope the first ship.
A 30-minute call. We'll map the highest-impact place to add AI in your stack and what it would take to ship.