Haelo
AI & Design

AI agents built a pre-sales design studio pipeline

Source
AI agents, one pipeline: how we built a pre-sales design studio
Published
Aniruddha Sainkar•25 July 2026

Aniruddha Sainkar, a product designer, describes a nine-agent pipeline built on Claude Code to compress pre-sales design work, from scouting SAM.gov opportunities through research, UI generation, QA audit and prototyping to a draft proposal.

The system runs on one architectural rule: a human gate sits between every stage, and no agent triggers the next.

The strictest constraint governs the generation agent, which must flag any pattern missing from the design system rather than invent a plausible variant, using Figma Console MCP, built by TJ Pitre, to read the file's actual components and tokens.

A separate QA agent then checks that output against the system for drift before a person ever opens the file.

Sources

Related articles

AI & Design

Using Claude and Figma MCP to automate UI assembly

Mai Sarah reports that automating UI assembly with Claude and Figma MCP requires a fully machine-readable design library first.

22 Sept 2026Mai Sarah · Stampede: the strategic design & technology company
AI & Design

AI Agents for Design: How Creative Teams Are Using Autonomous Tools in 2026

Kyros, whose own agent-orchestration tool gets pitched at the article's close, calls design system upkeep the AI agent "sleeper use case."

22 Sept 2026Kyros
AI & Design

8-agent pipeline that turns Figma designs into production-quality React design system components

Kaelig Deloumeau-Prigent built an 8-agent pipeline for Figma-to-code, with six deliberate human checkpoints baked in.

22 Sept 2026Kaelig Deloumeau-Prigent