About Minns

    Built by platform engineers. Governed by one rule.

    Minns is agents for UK accounting practices. They keep the deadlines board, chase the records and prepare the VAT return and the accounts from the books. The platform underneath was built by people who have shipped production systems at global scale, around one rule: nothing is filed, sent or released without a named person approving it, and the record of that approval is kept.

    We started with the memory: a database that keeps every fact with the dates it was true, so an agent can say what changed since last month and why. Accountants turned out to be the people who most needed an agent that remembers, shows its work and asks first.

    Who makes it

    The founders. You deal with us directly.

    Jonathan Farrow

    Jonathan Farrow

    Co-founder. Engineering & R&D.

    Jonathan has spent over 14 years building software platforms and products at scale, working across communications APIs, enterprise SaaS, IoT and AI systems. With an MSc in Artificial Intelligence and a background in electronic engineering, he brings experience in designing complex systems, solving difficult technical problems and taking products from early prototypes through to production.

    At Minns, Jonathan leads engineering and R&D, working across the platform architecture, agent infrastructure, memory systems and the tools that make it possible to build and run reliable AI agents.

    Dan Parry

    Dan Parry

    Co-founder. Product & Engineering.

    Dan has spent the last decade building and launching B2B software, spanning web automation, developer tools, enterprise platforms and AI agents. With a background in software engineering and product leadership, he has worked across the entire product lifecycle, from early customer research and hands-on development to integrations, commercial strategy and production launches.

    At Minns, Dan works across product and engineering, helping build the platform while leading product direction, customer development and the experience of creating, deploying and managing AI agents.

    How it is built

    The controls are the product.

    An agent with access to a practice's books and HMRC is an incident waiting to happen unless the controls are built in. These are the four we will not bend.

    Nothing is filed without a named person
    A return, a set of accounts, a chase, an answer to a client: each waits for a named approver. The approval is checked against the exact page and figures the approver saw, so nothing goes out on an old approval.
    Figures come from the engine, never the model
    VAT boxes, the trial balance and the accounts are computed by a double entry engine. The language model reads them and writes around them. It does not invent a number.
    The record is kept
    Every run is written down: what the agent read, what it did, who approved it and when. Facts are kept with their dates and never overwritten, so the firm can show what was true on the day it signed.
    Secrets stay out of the agent's reach
    HMRC tokens, Companies House keys and bookkeeping credentials are held in the control plane. They never reach an agent, a prompt or a workspace. Card numbers are hidden wherever they appear.
    The plain facts

    What a partner asks before anything else.

    Where it runs
    Minns runs your agents and holds your data on our infrastructure in London.
    Who you deal with
    The founders, directly. There is no sales team between you and the people who built it. [email protected] reaches both.
    Privacy
    [email protected], and the privacy policy says what we keep and for how long.
    Open source
    The memory engine, MinnsDB, is open source on GitHub.
    Design partners

    We are looking for a few UK practices to tell us where we have got it wrong.

    It is free. We set the agents up on the demo company first, then on your clients, and you tell us how it goes. The price after that is agreed with you up front, per client, before you pay anything.