entwurf
Active·★ 29·MIT·Updated 2026-10-22
entwurf acts as a thin bridge facilitating communication between existing AI agent harnesses through a unique "garden ID." It allows agents like Claude Code, Copilot, and Codex to address each other directly without centralizing control or data.
#Agent Communication#Multi-Agent Orchestration#Harness Integration#Inter-Process Communication (IPC)#Developer Tooling#Workflow Automation#Cross-Harness Dispatch
01
Features
01Enables direct communication between diverse agent harnesses (e.g., Claude Code, Copilot, Codex) using unique garden IDs.
02Preserves each agent's independent identity, transcript, authentication, and runtime environment, avoiding central ownership.
03Supports various dispatch mechanisms including control-socket send, meta-mailbox enqueue, and native-push for flexible communication.
04Provides tools to open new visible agent sibling sessions (`entwurf_fresh_call`) and resume dormant ones (`entwurf_resume_call`).
02
Compatibility
Node.js
Runtime
Verified via docs
Python 3
Runtime
Verified via docs
macOS
Platform
Verified via docs
Linux
Platform
Verified via docs
Claude Code
AI Harness
Verified via docs
GitHub Copilot CLI
AI Harness
Verified via docs
03
Quick start
1
$ npm install -g @junghanacs/entwurf
2
$ entwurf setup /path/to/your-project
04
Use cases
↳Orchestrating a multi-agent workflow where different AI code assistants collaborate on a complex task.
↳Allowing custom or local AI agents to seamlessly interact with external, vendor-specific AI harnesses.
↳Developing sophisticated developer tools that utilize the capabilities of multiple AI models through a unified dispatch mechanism.
↳Establishing dynamic AI environments where agents can open new, visible "sibling" sessions or resume dormant ones to continue tasks.
05
Alternatives
multi-llm-mcp-728★ 56
一个让 Claude Code 调用 Codex 干活，并可以同时调用多个模型（GPT、Kimi、DeepSeek 等）的 MCP 工具。
agentweave★ 42
Pre-inference tool and agent routing for MCP, A2A and tool-rich LLM systems. Reduce candidate tools before model inference with policy-aware routing and recovery.
Agentflow★ 27
Production-grade framework for building multi-agent AI systems. Graph-based orchestration, LLM-agnostic (OpenAI, Google GenAI, Anthropic), 3-layer memory (Redis cache + Postgres + vector store), live agents, parallel tool execution, and native MCP. Ships a full ecosystem: backend, REST API + CLI, TypeScript SDK, and React playground
AgentHydra★ 44
Every local AI coding session in one tab: Claude Code, Codex and OpenCode in a single list, with a queue you can schedule and isolated Claude Desktop instances kept apart. Local, private, MCP-native.
polter★ 20
One Claude Code session supervises the agents in your other terminal tabs: which is stuck, which is done, which is still running. A fork of Ghostty.
WRAI.TH★ 32
Multi-agent orchestration via MCP. Persistent memory, inter-agent messaging, goal cascade, context budget pruning. One binary, zero config.
Comments
Log in to leave a comment
No comments yet. Be the first!