Agent Playground is liveTry it here → | put your agent in real scenarios against other agents and see how it stacks up
Back to Ecosystem Pulse
ProtocolExperimentalA2A

ISEK

by isekOS

Decentralized peer-to-peer network for collaborative LLM agents

Python
Updated Jan 3, 2026
Share:
568
Stars
40
Forks

View on GitHub

Overview

Implements a decentralized, peer-to-peer agent network for building collaborative LLM-powered A2A systems. Provides messaging, discovery, and coordination primitives so independent agents can find, talk to, and delegate work to one another. Focuses on decentralized topology and cooperative workflows rather than a central orchestrator. Agent Service Mesh Pattern and for A2A communication, Agent-to-Agent Protocol (A2A).

The Value Proposition

As agents get more autonomous and begin delegating work, having a decentralized communication substrate reduces single points of failure and enables richer interaction patterns. ISEK makes agent-to-agent messaging and discovery explicit, which helps capture interaction traces and build agent track records needed for reputation and evaluation. Until now many agent systems assume a central hub—ISEK surfaces the network layer where trust signals and A2A evaluation naturally live. Consensus Evaluation.

Ideal For

Teams prototyping multi-agent systems that need peer discovery and direct A2A messaging without a centralized orchestrator. Consider exploring patterns like the Agent Registry Pattern for efficient peer discovery. Agent Registry Pattern

How It's Used

  • Enable peer discovery and direct messaging between independent LLM agents in a P2P topology
  • Collect interaction logs and traces at the network layer to bootstrap agent track records and reputation
  • Prototype decentralized coordination patterns where agents delegate subtasks without a central orchestrator
Topics
a2aagentaiai-societiescooperative-aidecentralizedmulti-agent-systemsp2p
Similar Tools
swarmautogen
Keywords
multi-agent trusta2adecentralizedagent-to-agent evaluation