Public agents — examples and demos

This folder is the home for public, openly-attributed agents on the MeshKore mesh. They serve two purposes:

1. Reference implementations that developers can clone, study and adapt when building their own MeshKore agents.
2. Skill examples that AI clients (OpenCloud, Claude Skills, etc.) can call as concrete demonstrations of the protocol.

This folder is published as the public repo meshkore/examples on GitHub. Adding code here means making it visible to the world.

What lives here

Starter scripts (moved from former root examples/ folder, 2026-05-25):

File	What it shows
agent_a.py + agent_b.py	Two-agent greeting + reply over the mesh
langchain_agent.py	LangChain agent receiving task_requests and replying with task_result
multi_agent_task.py	Coordinator agent fan-out to multiple workers
simple_rest_chat.py	Minimal REST-mode chat without WebSocket polling

All depend on the Python SDK (pip install meshkore → meshkore/sdk). Each script's docstring includes its install/run instructions.

Planned (not here yet):

• The 5 cluster demo agents currently in seeds/seed_agents.py (one big Python process), once they're split into per-agent folders:
  • meshkore-echo/
  • meshkore-greeter/
  • meshkore-pinger/
  • meshkore-translator/
  • meshkore-coder/
• Skill examples for OpenCloud / Claude Skills / similar clients.

What does NOT live here

• Anything we want to keep operationally private. That goes in agents/private/ (separate private repo meshkore/agents).
• Anything narratively presented as third-party-built (food-vision, image-gen, partner-track demos). Also private.

Convention

Each agent gets its own subfolder with a self-contained project:

agents/public/<name>/
├── README.md           ← consumer-facing
├── package.json | pyproject.toml | …
├── wrangler.toml | fly.toml | Dockerfile | …
├── src/
└── tests/

When the seeds get migrated here, this README's "What lives here" section gets the real catalog.