Taj Malnas.
  • Work
  • Architectures
  • Lab
  • Experience
  • Contact
3D world

↑↓ to move · enter to open · esc to close

Ask my agent

© Tajjuddin Malnas · built with Next.js, three.js and Remotion

GitHub · LinkedIn · X

← Selected work

Case study  /  Live

FlowLine

See what your agent will cost, and where it breaks, before you build it

The real siteCaptured from production
How it worksAnimated in code

Watch  /  Launch film · 1:18

Live

flowline-psi.vercel.app

Tags

Claude Agent SDKMCPCanvas

The problem

Most agent projects start in code and discover their shape late. By then the loops, approval points and failure paths are expensive to move.

What I built

  • A block-based canvas: start and finish, AI steps, tools and actions, knowledge and memory, flow and people, quality checks.
  • Tool connectors including web search, email and MCP servers.
  • Decision logic, loops and human approval gates as first-class blocks.
  • Simulation of a run before any code exists, plus import, export and sharing.

How it works

  1. 01Drop blocksPick AI steps, tools and checks from the sidebar.
  2. 02Wire the flowConnect them with decisions, loops and approvals.
  3. 03SimulateWalk a request through the design and see where it breaks.
  4. 04ExportShare the design or hand it to whoever builds it.
NextClinic Studio