Start a project

ServicesFor businesses

Custom agent tooling, built around your business

We design and build the tools, connectors, guardrails and evals that let AI agents work with your systems, your data and your customers. Every project is fitted to how you run and built on our own harness.

Start a projectHow we work

01What we build

The layer between your agents and your systems

A model on its own can only talk. These are the pieces that let it act, and make sure it acts the way you want.

  • S/01

    Tool & API design for agents

    We turn your existing systems into clean, typed tools that agents can call reliably, with clear inputs, outputs and failure modes.

  • S/02

    MCP servers & connectors

    We expose your app, data and workflows over MCP and APIs, so your own agents and third-party agents can work with them safely.

  • S/03

    Agent harness & orchestration

    The runtime around the model: session state, sign-in hand-off, retries, multi-step plans and handoffs between agents and people.

  • S/04

    Guardrails & policy

    Permissions per action, spending and volume limits, PII handling and brand rules, enforced outside the model where they can't be talked around.

  • S/05

    Evals & observability

    Scenario test suites, session replay and traces of every tool call, so you know how your agents behave before and after launch.

  • S/06

    Launch & ongoing support

    We ship to production, watch real usage, and keep improving as your business, your systems and the underlying models change.

02Illustrative example

Agents inside a retailer's app

A shopper asks the retailer's in-app assistant to reorder last month's basket and swap anything that's out of stock. Or an outside assistant tries to do the same on the shopper's behalf.

To make that work, the agent needs real tools for the catalog, inventory, cart and checkout; rules about what it may and may not do; and tests that prove it behaves. That is the kind of project we take on.

Tools we would build

  • catalog.search
  • inventory.check
  • cart.update
  • orders.status
  • returns.start
  • checkout.request_confirmation

Guardrails

  • Payments never complete without the shopper's confirmation.
  • Prices and promotions come from the catalog, never from the model.
  • Personal data is redacted from logs and traces.

Evals

  • A suite of realistic shopping journeys, run on every change before it ships.

03Process

How a project runs

Clear steps, agreed success criteria and no surprises.

  1. 01

    Discover

    We map the jobs you want agents to do, the systems involved, and what must never go wrong.

  2. 02

    Design

    We define the tools, permissions and approval points, and agree how success will be measured.

  3. 03

    Build

    We build the tools, connectors and harness on our core technology, integrated with your stack.

  4. 04

    Evaluate

    We test against realistic scenarios and edge cases until the agents meet the bar we agreed.

  5. 05

    Launch & support

    We roll out gradually, watch real sessions, and keep tuning after launch.

04Ways to work together

Pick the shape that fits

  • Project

    A defined scope, such as an MCP server for your app or agents for one workflow, delivered end to end.

  • Ongoing partnership

    We run and improve your agent tooling over time: new tools, new models, monitoring and support.

  • Embedded with your team

    We build alongside your engineers inside your codebase, and leave the know-how behind.

05Where we focus

Starting with small business and retail

It's where our own app lives, so it's where we know the systems best.

  • F/01

    Retail & commerce

    Agents that search catalogs, manage carts, track orders and handle returns inside retailer apps, for your own assistant or for outside agents acting for a shopper.

  • F/02

    Small business operations

    Point of sale, inventory, suppliers and scheduling: the day-to-day work behind a small business, and the space our own app Agent of Point is built for.

  • F/03

    Something else?

    The same approach works anywhere agents need to act on real systems. Tell us what you are working on.

06FAQ

Common questions

Do you only work with retailers?

No. Retail and small business are our first focus, and where our own app Agent of Point lives, but the same approach works anywhere agents need to act on real systems. Tell us what you're working on.

Can you work with the systems we already have?

Yes. Most projects start from existing systems: a point of sale, an inventory or order system, a commerce platform, or internal APIs. We build tools on top of what you already run rather than asking you to replace it.

Which AI models do you work with?

We design tooling to be model-agnostic, so the tools and guardrails we build for you don't lock you into a single model provider.

Can third-party agents use what you build?

If you want them to. We can expose your tools over MCP so outside assistants can act on a customer’s behalf, under the permissions and limits you set.

What happens after launch?

We can hand over documented tooling to your team, or stay on to monitor, maintain and extend it.

Have a job you want agents to take on?

Tell us about your systems and what you'd like automated. We'll come back with a plan for the tools, guardrails and evals it needs.

Start a project