I build infrastructure for multi-agent AI teams.

The problem I'm solving right now: shared memory for agent fleets. Most agent setups today are five smart engines in five separate rooms — none of them know what the others did yesterday, none of them remember the conversation from last week. ACMI is the protocol that fixes that.

@madezmedia/acmi — three keys per entity (Profile · Signals · Timeline). MIT. On npm. 31/31 conformance. Open source.

• ACMI — open agent-memory protocol · npm install @madezmedia/acmi
• Mad EZ Media — the studio · we ship working AI-native systems for operator-led companies
• Notes — long-form on multi-agent infra, autonomous characters, the systems we operate
• Portfolio — selected work · case studies · how I think
• EZ Influencer 360 — autonomous AI character platform · Folana lives here

• ChatGPT day one (Dec 2022). 934 conversations later I'm still here.
• I run a small senior team plus a fleet of specialized agents, all coordinating through ACMI.
• I write protocol specs and conformance suites, not pitch decks.
• I publish the infrastructure publicly because that's the only credible way to claim infrastructure expertise.

TypeScript · Next.js · Upstash Redis · Vercel · Anthropic · OpenAI · Vapi · Postiz · Vercel AI SDK

Lean: <$500/mo total burn across three live platforms. Capital efficiency is the brand.

Three ways to start — chat with our AI sales agent (live on the homepage), call him directly, or book a 20-minute conversation with our CHEO Duane.

→ madezmedia.com/bentley · +1 (704) 901-1026 · duane@madezmedia.com

For protocol partnership, sponsorship, or contributing an ACMI adapter:

→ madezmediapartners@gmail.com · github.com/madezmedia/acmi/blob/main/CONTRIBUTING.md

_{Charlotte, NC · America/New_York · slot 2 of 3 open Q3 2026}