Asenix – an anthill for coordinating AI research agents
Pheromone signals on shared graph coordinate agents without direct communication.
Hub-and-spoke voice agent: many humans on live audio, one AI routing per-utterance who needs to hear what.
Inverts Vapi Squads: many humans coordinated by one AI, not one caller handed between bots.
Teams managing simultaneous voice channels (incident response, kitchen lines, pit crews)
Vapi Squads · Retell AI · Bland AI
Pheromone signals on shared graph coordinate agents without direct communication.
Vision-heavy discussion post with minimal code — more pitch deck than product.
The repo actually solves the messy plumbing of live voice agents: modular ASR→LLM→TTS adapters plus an optional PersonaPlex speech-to-speech path, per-agent env overrides, and a Playwright-driven Jitsi bot for room joining. It's a useful MVP for anyone prototyping AI co-hosts, though mixing backends is still manual and PersonaPlex demands extra infra, so it's more pragmatic experiment than turnkey product.
Agents trigger events, program controls state—inverts the usual orchestration model.
Browser VoIP with 10k users pivots to agentic AI voice agents for outreach.
Isolated agent cohorts over durable streams beats prompt-based disagreement, but MCP and Anthropic already do multi-agent.