I built a toolkit for building hive minds with AI agents (github.com)

A developer unveiled Ecco, an open-source SDK for building “hive minds” of AI agents — a peer-to-peer mesh that lets local agents discover, authenticate, negotiate and act on users’ behalf across both public and air-gapped networks. Ecco combines zero-config mDNS for local swarms, a DHT for decentralized global discovery, and gossip-based pub/sub for capability broadcasting and state updates. Each node holds a cryptographic identity (public/private keypair) and advertises structured “capabilities” that a matcher uses to pair clients with suitable agents. Optional centralized registries (hono + Postgres + Redis) can provide global visibility, reputations, monitoring and analytics while remaining optional for privacy-minded or isolated deployments. Technically the SDK also provides flexible orchestration patterns: selection strategies (all, top-n, round-robin, weighted, random) and aggregation strategies (majority/weighted vote, best-score, ensemble, consensus-threshold, first-response, custom) to coordinate multi-agent responses and reach consensus. The approach targets scalable, interoperable ecosystems where personal agents coordinate with business agents autonomously, while retaining cryptographic trust and options for local-only meshes. Significance for AI/ML: Ecco makes decentralized multi-agent coordination practical, enabling new architectures for privacy-preserving automation, federated workflows and robust ensemble decision-making — though it also raises governance, safety and trust-management questions as such networks scale.