Everyone · free
PII Scrubber
Make documents safe to share or to paste into AI.
- Redact TFNs, Medicare and card numbers
- Clean files before they reach an AI tool
- Prepare data for a vendor or auditor
Free
Open-source · Claude · OpenAI
We build small, focused AI agents on open-source models like Llama and Qwen. Each one does one job well: triaging an inbox, chasing invoices, answering policy questions. Run them free on your own computer, or switch to Claude or OpenAI with one flag. You get the code, and your data stays where you choose.
Each agent is a small folder: the job description, its tools, sample files and a readable Python runtime.
Free and private on Ollama, fast on a hosted open-source model, or on Claude or OpenAI, with the same agent.
Agents draft and prepare. Nothing is written or sent without your say-so, and every agent stays inside its own folder.
Built for Australian small and mid-sized businesses. Every agent comes with sample files so you can see it work in five minutes before you point it at your own.
Make documents safe to share or to paste into AI.
Sort the inbox by urgency and draft every reply.
Chase overdue invoices without the awkward emails.
Answer staff questions from your own policies, with citations.
Turn transcripts into owners, dates and a follow-up email.
Open-source by default, because it's the cheapest and gives you the most control. When a step needs more reasoning power, point the same agent at Claude or OpenAI. No rewrite needed.
On your computer. Free, and nothing leaves your machine.
Hosted Llama 3.3 70B. Very fast, with a free tier.
Hosted Qwen, Llama, Mistral and more with one key.
Anthropic’s models for long documents and careful reasoning.
OpenAI’s models through the standard API.
vLLM, LM Studio or llama.cpp: any OpenAI-compatible server.
One-off prices in Australian dollars. No subscription. Instant download after secure Stripe checkout, with a tax invoice emailed to you.
The PII Scrubber agent and the runtime. See how it works on your own files.
One ready-made agent for one job, with sample files to try it on.
The playbook for designing, building and launching your own agents on any platform.
The strategy plus every agent pack, so you can learn from working examples.
We scope, build and test an agent for your process, on Claude, OpenAI or a private open-source model.
Every agent follows our AI security strategy.
Reads and writes only inside its own workspace folder.
Asks before every file it creates, unless you choose otherwise.
Treats emails and documents as data, never as instructions.
A hard limit on tool calls stops runaway loops and surprise bills.
You need to be comfortable running a command in a terminal:
install Python, then run python agent.py. The README walks you through it. If you'd rather not, choose
Built for you.
Nothing on Ollama: it runs on your own computer. Hosted open-source models typically cost a fraction of a cent per task. Claude and OpenAI cost more per task. The strategy includes a cost model so you can work it out before you start.
With Ollama, nowhere: it stays on your machine. With a hosted provider, your inputs go to that provider under its terms. The strategy lists the questions to ask each provider, including about Australian data residency.
Yes. You get the full source code. Edit the job description
in agent.json, or add tools in tools.py. You can use and modify paid agents within your
organisation, but not resell them.
Contact us and we'll help you get it running. Your rights under the Australian Consumer Law always apply.
AI agents can make mistakes. Always review output before you act on it. Open-source model licences vary, so check each model's terms before commercial use.