Watch the builds.

18 short films showing how the systems work, from the problem to the run to the result. Each one is under a minute.

The films use sample data, so the figures on screen illustrate how a system behaves. They are not client results.

Claude Code

See the service
  • Developer tooling

    Accessibility, fixed at the source

    Claude Code crawls every route with Playwright, traces hundreds of accessibility violations back to the few shared components causing them, and opens one pull request that fixes the source code rather than patching pages.

  • Developer tooling

    API docs that fail the build

    When a route handler changes, the spec, docs, changelog and SDKs update together, and every documented example runs as a contract test. Wrong docs become a red build, not a support ticket.

  • Database performance

    Database queries, ranked by cost

    Queries are ranked by total cost rather than how slow they feel, traced back to the exact line of code, and every proposed fix is benchmarked on a restored snapshot before it reaches production.

  • Claude Skills

    Six languages, shipped on time

    A Claude Skill translates each release's changed strings into six languages, enforces the glossary and placeholders, and holds back anything that breaks a rule for a human to check.

  • Claude Cowork

    Onboarding, ready before Monday

    Claude Cowork drafts the first-week plan, IT checklist and manager brief from your own templates, then checks HR and IT tickets daily so every open item has an owner. Drafts wait for review; nothing is sent automatically.

  • Claude Skills

    A reusable Claude Skill for business workflows

    A repeatable business process packaged as a Claude Skill: structured instructions plus an execution system, so Claude runs it the same way every time and the Skill improves from each result.

Websites and apps

See the service
  • Web app

    A full-stack app with Claude inside

    One web application with interface, backend logic and stored data, where Claude makes the judgement calls inside the product and the app owns the state, rules and records.

  • Client portal

    A customer portal that already knows them

    Clients sign in and ask questions in plain language, and Claude answers from their own account data, keeping context for follow-up questions. Built with Claude, Claude Code, APIs and authentication.

  • Finance operations

    Three-way invoice matching, done in code

    Invoices from the accounts inbox are read to a strict schema, matched against purchase orders and goods receipts, and every total is recomputed in code. Clean matches queue for approval and only the exceptions reach a reviewer.

  • Competitive intelligence

    A competitor price monitor that fails loudly

    Playwright collects competitor prices and validates each one against history. When a scraper silently breaks, it raises an alert instead of storing bad data, so decisions never rest on stale prices.

  • Purchasing

    Restock planning: arithmetic first, judgement after

    Reorder points and quantities are computed in code from sales history, seasonality and supplier lead times. Claude then adds buyer notes and supplier context to a weekly brief grouped by supplier.

  • Compliance

    Compliance evidence, collected as it happens

    Read-only collectors gather audit evidence across identity, cloud, code and ticketing as it happens, check it against each control's wording, and flag gaps inside the quarter instead of at audit time.

  • Sales operations

    An AI proposal generator

    A project brief goes in and a structured proposal draft comes out in your wording and order: understanding, scope of work, approach, timeline and price from your own rate card, and next steps.

  • Operations

    Every inbox thread gets an owner

    An AI agent reads whole threads, classifies them, pulls out order numbers and dates, and routes each to the right team. Low-confidence threads go to a review lane, and every decision is logged with its reasoning.

  • AI agent

    An AI agent that plans, reasons and executes

    A web app where users hand over complex tasks and a Claude agent plans the steps, calls tools and reports back, with approval gates, full step visibility and the ability to stop and redirect a run.

  • AI agent

    One request, many systems

    An agent turns a plain-language request into a plan and works across the tools you have approved. Claude decides what should happen, a separate layer carries it out, and it can only reach the scopes you grant.

  • AI quality

    Quality assurance for AI features

    A test suite for AI features that judges answers against rules instead of exact wording, sorts every failure into a named type, and re-runs fixed cases on every build so the same issue cannot come back unseen.

MCP servers

See the service
  • MCP server

    A 2009 ERP, spoken to plainly

    A typed MCP server gives Claude live access to a legacy ERP: products, stock levels, customer pricing and open orders, with the one write action, drafting a quote, held behind a confirmation and an audit log.

Start a project
Denver Emerald Peter

Want one of these for your business?

Tell me what eats your week. You get a plan and a fixed price within 24 hours.

Start a project