Agents that actually do the work
Four agents built for specific jobs, with their architecture, tools and execution loops laid out. If you want to see what a working agent actually looks like underneath, this is it opened up.
Spatial UI Synthesizer Agent
2D Spec & Figma to 60 FPS Three.js Spatial Scenes
Fuses mathematical camera frustum coordinate translation with Emil Kowalski spring mechanics and WebGL GPU performance budgets.
Telephony Voice Bridge Agent
Sub-Second Conversational Speech over Telephony Streams
Bi-directional 8kHz mu-law audio chunk pipeline with immediate caller barge-in buffer clearing and ambient filler loop.
Autonomous Quality Guard & Regression Hunter
Git Worktree Swarm Isolation & Keep/Reject Test Gates
Dispatches worker agents into isolated git branches, enforces fail-closed test suites, and rejects hallucinations before merging champion commits.
Answer Engine Optimizer Agent (AEO)
Machine-Readable Discovery & Citability for LLMs
Parses website entity graphs, aligns crawler robots policies for AI bots, and exports verified llms.txt and ai-index.json manifests.
Machine Layer & Standard JSON Resume
All technical metrics, active agent runtimes, and skills are exposed through standard JSON Resume specification for programmatic consumption by LLM coding agents.
curl -s https://mrmp.pro/api/resume.json | jq .basics