The platform

One platform. Four ways to put it to work.

Everything we sell runs on the same engine. You buy at whichever level fits — from a single-task assistant to a full process redesign — and it all shares one account, one login, one bill. The more of it you use, the more it knows about your business.

Take the AI Readiness Check Skip to the technical bit
The platform at a glance
Ways to use it Four levels
Account One per organisation
AI models Eight providers, switchable
Your data Yours, walled off
Every product we sell runs on this same engine.
What you're buying

Four levels. Start at any of them.

Each level is a different amount of "done for you". They stack — plenty of customers use two or three at once.

1

Assistants

Lightest

Simple, single-shot tools. Give one a brief and it hands back a finished file — a slide deck, an image, an infographic, a diagram — styled to your brand. The fastest way to feel what agents can do.

See the assistants
2

Build your own

DIY agents

Build your own agents through a guided conversation — the Agent Architect turns your idea into a working agent, with your fields, your choice of AI model and your outputs, and no code anywhere. Single-shot agents today, with richer dynamic agents on the way.

Building on the platform
3

MAF applications

Products

Complete workflow products that take a whole job end to end — Meeting Assistant for conversations, Menu Magic for hospitality, and more arriving at a rapid rate.

See the products
4

Working with us

Done for you

Projects and integrations, built on AI. We redesign one of your processes around agents, wire it into the systems you already run, deploy it, and iterate — a structured engagement, or a fully bespoke build when the shape demands it.

How we work
Under the hood

Our own platform — and why that matters to you.

We built the engine ourselves, and everything we sell runs on it. That's not tech vanity — it's what makes the products reliable, auditable, and secure, and it's why nothing here locks you to one AI vendor.

  • Reliable orchestration — a standard way to run agents, with cost tracking and memory logging on every call. You can see what ran, what it cost, and what it produced.
  • Multi-layered memory — agents keep hold of client histories, house rules and past decisions, backed by database storage for structured data and file storage for the artefacts they produce. The business gets sharper the more it runs.
  • Prompt management & coordination — prompts are managed and A/B tested rather than buried in code, and multi-agent coordination handles the jobs that take more than one specialist.
  • Eight AI providers, switchable — OpenAI, Anthropic Claude, Google Vertex, DeepSeek, Perplexity, Ollama, RunPod and Fal.ai, chosen per agent or per stage. You're never locked to one vendor — or one price curve.
  • Real-world tools — Xero, Companies House, Google Maps, WordPress, LinkedIn, transcription and CRM integrations. Agents act on real systems, not just chat.
  • 190+ built-in utilities — file handling, PDF generation, web scraping and more, so new agents get to the real work faster.
  • Security is paramount — one account per organisation, user entitlements, and row-level security. Your data stays walled off, cleanly.
architecture
orchestration enginerun_agent · run_tool_agent
├ cost trackingper call
├ memory loggingautomatic
└ error handlingbuilt-in
memoryworking → canonical → vector
storagedatabase · files & artefacts
promptsmanaged · A/B tested
coordinationmulti-agent
toolsXero · Maps · CRM · web
identitytenant · user · RLS
modelsOpenAI · Claude · Vertex · DeepSeek · Perplexity · Ollama · RunPod · Fal.ai
utilities190+

Every operation is logged with what it did, what it cost, and where the data came from — so when an agent acts on your behalf, you can always answer "why did it do that?"

See it against your work

The platform is real. Let's point it at a process of yours.

Tell us what you're trying to change, and we'll show you how the levels come together around one of your workflows.

Talk to us