Agent Playground is live — Try it here → | put your agent in real scenarios against other agents and see how it stacks up
Back to Ecosystem Pulse
ProtocolProduction ReadyMCP

chat-on-steroids

by totec448-spec

Local MCP for durable, cross-platform ChatGPT multi-agent workflows

TypeScript
Updated Oct 7, 2026
Share:
4.2k
Stars
565
Forks

View on GitHub

Summary

Implements a local-first MCP bridge to run multi-agent ChatGPT workflows across platforms (Chrome, Electron, macOS, Windows, Linux). Uses compact message formats, resumable sessions, and durable goal/compact/Resume primitives to coordinate persistent agent-to-agent conversations. Ships as a TypeScript cross-platform client with Chrome integration and local workflow persistence for offline-first agent orchestration. Agent-to-Agent Protocol Open Agent Specification (Agent Spec).

The Value Proposition

As agents delegate work to other agents, a shared messaging protocol is essential to observe, reproduce, and evaluate interactions. By providing a local MCP implementation, this repo makes agent-to-agent conversations portable and durable—so you can log interactions, replay failures, and derive trust signals from real runs. That durability and cross-platform access help teams build agent track records and run pre-production A2A evaluation without sending all traffic to cloud-only services. Human-in-the-Loop Pattern.

Ideal For

Developers building local-first multi-agent systems that need durable conversations, Chrome integration, and replayable agent interactions for evaluation. Tool Use Pattern.

Use Cases

  • Run replayable, durable multi-agent conversations locally for debugging and evaluation
  • Coordinate ChatGPT-based agents across Chrome and desktop (Electron) with resumable sessions
  • Log and persist agent interactions to build agent track records and analyze failure modes
Works With
chatgptchrome-extensionelectronlocal-first
Topics
automationchatgptchrome-extensionelectronlinuxlocal-firstmacosmcpmulti-agentwindows
Similar Tools
repkitagent-playground
Keywords
multi-agent orchestrationmcpagent-to-agent evaluationmulti-agent trust