interfaces for humans

Work with us

Ship generative UI that does what it shows.

We help product, platform and design-system teams put AI-composed interfaces and agent actions in front of real users: the right rung on the spectrum, the right protocol, consent before every consequential effect, and proof that it works.

Who calls us

  • Platform teams shipping an app into ChatGPT, Claude or another agent host
  • Product teams adding a copilot that can take actions
  • Design-system teams making components agent-operable
  • Risk owners in fintech, health and enterprise SaaS who need evidence
  1. 015 business days

    Gen-UI Teardown

    One agent-facing flow you're worried about.

    • RUI run in a real browser on the flow
    • 25-minute recorded walkthrough
    • Five-page findings: what it does vs what it shows
    from $1,500
  2. 021–2 weeks

    App Store Pre-submission Audit

    Teams about to submit an MCP App to ChatGPT, Claude, Microsoft 365 Copilot or Gemini.

    • Your app run in each host you target
    • Consent, CSP, display modes and accessibility checked against each review gate
    • Fix list ranked by rejection risk
    • Evidence bundle you can hand to reviewers
    from $5,000
  3. 03most teams start here

    Readiness Audit

    Teams about to put generated UI or agent actions in front of customers.

    • Surface map: in-app copilot, MCP Apps, AG-UI, A2UI
    • Behavioural tests on up to five flows
    • Consent and reversibility classification for every action
    • Risk register, 90-day roadmap, exec readout
    from $9,0002 weeks
  4. 04half or full day

    Team Workshop

    Engineering and design teams who need a shared mental model, fast.

    • “Generative UI from the wire up” using this guide's labs
    • Hands-on build with Agentable UI
    • Up to 25 people, remote or on-site
    from $6,000
  5. 054–6 weeks

    Protocol Integration Sprint

    Shipping one production agent surface properly.

    • MCP App, AG-UI or A2UI surface in production
    • Consent gates and reversibility metadata
    • Behavioural checks in CI
    • Handoff docs your team owns
    scoped · typically $30–75k
  6. 066–10 weeks

    Design System for Agents

    Design-system teams told to “make our components work with AI”.

    • Component capability manifests and schemas
    • Accessibility-tree hardening
    • A generation catalog and composition rules
    • Approved-component registry
    scoped
  7. 073-month minimum

    Advisory

    Leaders who want a second brain on agent-UI architecture.

    • Monthly architecture review
    • Async access in your Slack
    • Quarterly protocol and market briefing
    from $6,000 / month

Fixed scope, fixed price.

You know the cost before we start. No hourly meters.

Evidence over opinion.

Findings cite what a real browser observed, not what a screenshot implied.

Protocol-neutral.

We build on AG-UI, A2UI, MCP Apps and WebMCP. We don't sell you a protocol.

You own the result.

Code, docs and checks live in your repo. We don't become a dependency.

Start with a 30-minute call.

Tell us what you're building and what worries you. If we're not the right fit, we'll say so and point you to who is.

Email hypharsh@gmail.com →DM on XOr learn it yourself →

RUI findings are evidence for your team's decisions, not a certification. We publish anonymised patterns from engagements only with permission.