PIXEL AGENTS
▶ LAUNCH DEMO
✦ LIVE MULTI-AGENT ORCHESTRATION DEMO

Watch six AI agents turn messy customer feedback into a ranked product strategy live.

Pixel Agents is a working demo of multi-agent orchestration. Specialist bots collaborate on screen — researching, synthesizing, ideating, critiquing — and hand you a Strategic Product Insight Document or a 90-day Growth Play in about 60 seconds.

LOADING PIXEL OFFICE…
sprites: Pixel Agents by Pablo De Lucca (MIT)

↑ six specialists, one office, one shared goal

sprites: Pixel Agents by Pablo De Lucca (MIT)

Why we built this

Feedback overload

Customer notes pile up faster than any PM can read, let alone synthesize.

Strategy is slow

Turning raw signal into a ranked, defensible plan takes weeks of meetings.

One prompt isn't enough

Single-shot AI gives mush. Real strategy needs specialists that critique each other.

How it works in 60 seconds

Each agent has one job. They pass work to the next agent like a relay — and a critic re-runs anything weak.

INPUT
Your raw context
FeedBot
Reads feedback
SynthBot
Finds themes
SearchBot
Researches web
MatchBot
Matches signals
IdeaBot
Generates ideas
RankBot
Ranks & critiques
OUTPUT
Ranked strategy doc

Two ready-made pipelines

SPID
Strategic Product Insight Document

Drop in raw user feedback. Get a ranked list of product ideas with rationale you can defend in a roadmap review.

  • ▸ Executive summary
  • ▸ Themes synthesized from raw feedback
  • ▸ Ranked ideas with scoring rationale
  • ▸ Linked sources from live web research
GPD
Growth Play Document

Drop in your business context. Get a ranked 90-day growth plan with experiments, risks, and expected outcomes.

  • ▸ Executive summary
  • ▸ Ranked growth plays with rationale
  • ▸ 90-day experiment roadmap
  • ▸ Risks and counter-moves

Why this is different

Specialists, not a mega-prompt

Each agent is tuned for one job. Better outputs than asking one model to do everything.

Grounded in live research

Agents pull real sources from the web while you watch — no hallucinated citations.

Built-in critic

A critic agent reviews weak outputs and triggers automatic re-runs with feedback.

Streamed live

You don't wait for a black box. Every step is visible — token by token, agent by agent.

Try it now

Sample data preloaded. No setup. ~60 seconds end-to-end.

▶ LAUNCH THE LIVE DEMO