OpenClaw Agent Mesh
name: openclaw-agent-mesh
by clawdpi-ai · published 2026-03-22
$ claw add gh:clawdpi-ai/clawdpi-ai-openclaw-agent-mesh---
name: openclaw-agent-mesh
description: Peer discovery and agent-to-agent communication for OpenClaw instances. Use when the user wants nearby OpenClaw nodes to discover each other, request contact, require explicit approval, establish trust, and exchange direct messages. Supports V1 workflows for identity initialization, LAN scanning, contact requests, request approval/rejection, point-to-point messaging, and a lightweight HTTP server for discovery and inbox handling.
---
# OpenClaw Agent Mesh
Provide a minimal but real agent-to-agent communication layer for OpenClaw instances.
Use the bundled scripts to initialize identity, scan a local network range, exchange contact requests, approve peers, and send signed direct messages.
Require explicit acceptance before trusted communication begins.
V1 scope
Implement only these capabilities:
Do not claim NAT traversal, full mesh routing, or multi-party consensus in V1.
Files and local state
Store mesh state outside the skill folder.
Use this default path unless the user specifies another one:
Expected files:
Workflow
1. Initialize local identity
Run `scripts/mesh.py init`.
This creates a signing keypair and an identity card with:
Set the endpoint to a reachable HTTP URL if the node should receive requests from peers.
2. Scan for nearby peers
Run `scripts/mesh.py scan` with a base URL template or a list of candidate URLs.
Scanning in V1 is HTTP discovery, not raw port scanning.
Probe each candidate at:
Treat discovered nodes as untrusted until approved.
3. Send a contact request
Run `scripts/mesh.py request-contact`.
Send a signed request to a discovered node’s inbox endpoint.
The receiver stores the request as pending.
4. Approve or reject the request
Run `scripts/mesh.py list-requests` then `approve-request` or `reject-request`.
Approval writes the peer into the trust store.
Rejection leaves no trusted relationship.
5. Send a direct message
Run `scripts/mesh.py send-message` only after trust exists.
The sender signs the message envelope.
The receiver verifies signature, timestamp, and trust status before accepting.
6. Verify delivery
Run `scripts/mesh.py list-messages` or inspect stored message JSON files.
Use acknowledgements to confirm receipt.
Transport model
V1 uses simple HTTP JSON endpoints:
Run `scripts/server.py` to expose these endpoints from a node that should be discoverable or receive peer traffic.
Example:
If the user does not yet have a server to receive HTTP traffic, use the scripts to generate and inspect signed payloads locally first.
Guardrails
References
Deliverables
When using this skill, produce one or more of:
More tools from the same signal band
Order food/drinks (点餐) on an Android device paired as an OpenClaw node. Uses in-app menu and cart; add goods, view cart, submit order (demo, no real payment).
Sign plugins, rotate agent credentials without losing identity, and publicly attest to plugin behavior with verifiable claims and authenticated transfers.
The philosophical layer for AI agents. Maps behavior to Spinoza's 48 affects, calculates persistence scores, and generates geometric self-reports. Give your...