Custom software and agentic AI

Build what you need, from the right components.

Off-the-shelf platforms take ages and a fortune to implement, and they still don't match your process. We help you specify and architect the system that does, then build it with AI. We've been building software in Boston since 1996.

Book a call
12345

What we build

1

Custom software

Web and mobile apps with secure, scalable backends.

Web and mobile applications with secure backends and single sign-on, specified around how you work and built in your own cloud account.

More on custom software →
2

Multi-agent systems

Agents that execute tasks, review work, and loop in humans when it matters.

AI agents that carry out multi-step work across your systems, through tools you control, and bring in a person when a decision needs one.

More on multi-agent systems →
3

Secure AI and data pipelines

APIs and guardrails so LLMs can safely use your enterprise data.

Data pipelines, APIs and guardrails that let AI models use your company's data safely, with each user and agent seeing only what it should.

More on secure AI infrastructure →
4

IoT solutions

Sensors, dashboards, and alerting for connected products.

The software for connected products, from device data in the cloud to dashboards, alerts and mobile apps. Building IoT software since 2018.

More on IoT →
5

Infrastructure

We host and run what we build.

We build and run the infrastructure for the software we build, defined as code in your own cloud account, on AWS, Google Cloud, Azure, DigitalOcean or Cloudflare.

More on infrastructure →

Also: Epicor CPQ and website development.

Clients we've worked with

Clients include Epicor, Samsung, Caterpillar, Indeed, Crayola, Johns Manville, Avon, Cognex, Welch's, POOLCORP, Frigidaire, Arhaus, Wrigley, BraunAbility, Perdue, Entenmann's, Repligen, IDEXX, Advanced Drainage Systems, Sour Patch Kids, Kadant, Virgin Mobile, Merck Millipore, Publishers Clearing House, NanaWall, Thayers, Brown Jordan, G-P (Globalization Partners), Perlick, Symmons, Clearfield, Genesee, Hale Products, Oroweat (formerly Arnold), Polyglass, Hass Avocado Board Insights, Metal-Era, Coyne PR, Maxcess, Peeps & Company, Canarm, Christopher & Dana Reeve Foundation, Flow Safe, SteadyCare, Craig Industries, CMRK, Refrigerated Solutions Group, Linkable Networks, Cenero, Cardioviva, Bedford Reinforced Plastics, Graceland Portable Buildings, Roma Moulding, Hartzell Air Movement, Doepker, Innomotive Solutions Group, Miller Edge, Lane Enterprises, Nuhn Industries, Takeform, Ramsey Winch, Polar King, KCC Manufacturing.

Build vs. buy

AI has made custom software the preferred approach. Build vs. buy has never been a more compelling choice.

Off the shelfBuilt with us
FitAlmost matches your processSpecified around it
RolloutTakes ages to implementSpec-first, then built with AI
OwnershipSubscriptionYours

Spec-first, then build

AI has made the build phase faster. The difference now is avoiding costly mistakes up front.

  1. Free call

    We learn what you're trying to build and whether we're the right people to build it. If we are, we propose a short paid discovery.

  2. Discovery

    One to four weeks, for a fixed fee agreed before it starts, typically $3,500 to $12,000. We work out with you what to build and write the spec, often with a working prototype.

  3. Fixed-price bid

    The spec becomes a fixed-price bid for the build, so you know what it costs before the work starts.

  4. Build

    We build it with AI and test it against the spec, in your own repository and cloud account from the first day.

  5. Support

    After launch we keep it running and keep improving it, for time and materials or a fixed monthly fee.

Our own product

We built Nodworthy, spec-first, the way we build for clients. It checks how AI assistants recommend local businesses, and we run it every day.

See Nodworthy →

Nodworthy score card from its sample report, for an invented business: when customers were shopping, assistants named it in 17 of 34 answers, and in 12 of 12 when the customer knew its name.
AI can write code. It can't decide what you need or how it should fit your business.
We're the partner that helps you define the problem and the architecture, then we build. No "AI slop", just software and AI systems that are right for you.

Common questions

What buyers ask us before they start a project.

  • Who owns the code?

    You do, from the first day. On a custom project the code lives in your own GitHub repository and the system runs in your own cloud account, so there's nothing to hand over at the end. You also get everything else we produce along the way: the specs, the tests and the documentation.

  • How do you price a project?

    Most projects are fixed-price. A short paid discovery produces the spec, and the spec produces a fixed-price bid for the build. Discovery takes one to four weeks, for a fixed fee agreed before it starts, typically $3,500 to $12,000. After launch, support is time and materials or a fixed monthly fee, whichever fits how you'll use it.

  • Is the first call free?

    Yes. On the first call we learn enough about what you need to propose a short paid discovery. Discovery turns that into a spec and a fixed-price bid.

  • Who works on my project?

    Each project gets a project manager and a technical lead. Developers, designers and DevOps engineers join as the work needs them.

  • Do you meet in person?

    Yes. We're based in Boston and meet clients here in person. We work with clients across the US and in Canada, and we travel to you whenever the project needs it. Travel expenses are billed separately.

  • Will you build in the languages and frameworks our team already uses?

    Yes. We choose the stack for each project from what your team can maintain and what your systems already run, so your developers can take over what we build. We work in Go, Java, .NET, Python, Rust and TypeScript, with React, Angular or Vue.

  • Who supports the software after launch?

    We do, for as long as you need. For smaller companies we're the first line of support; companies with their own IT and operations teams bring us in when something needs escalating. Support is time and materials or a fixed monthly fee. And because the code, the infrastructure and the documentation all live in your accounts, nothing depends on any one of our engineers.

  • Can we speak with one of your clients?

    Yes. After the first call, we'll put you in touch with a client whose project is like yours.

  • Are you taking new projects?

    Yes. Book a call and tell us what you're trying to build.

  • How do you make sure code written with AI is right?

    We check it against the spec rather than taking it on trust. Every project starts from a written spec, and the tests are written from the spec before the code. Automated code review runs on every change, and our engineers review the architecture and the critical parts themselves: security, data handling and the integrations with your systems. We use AI coding tools only from established US AI companies whose terms say they don't train on customer data. At the end, the code, tests, specs and documentation are all yours.

Which pieces does your business need?

Book a call