OpenSkillIndex← back to search
claude-code

Asi tailscale-localsend

Tailscale + LocalSend Peer Discovery

install
source · Clone the upstream repo
git clone https://github.com/plurigrid/asi
Claude Code · Install into ~/.claude/skills/
T=$(mktemp -d) && git clone --depth=1 https://github.com/plurigrid/asi "$T" && mkdir -p ~/.claude/skills && cp -r "$T/skills/tailscale-localsend" ~/.claude/skills/plurigrid-asi-tailscale-localsend-30d7cd && rm -rf "$T"
manifest: skills/tailscale-localsend/SKILL.md
source content

Tailscale + LocalSend Peer Discovery

Discover peers via Tailscale mesh and exchange files via LocalSend protocol.

Architecture

┌─────────────────┐     ┌──────────────────┐     ┌─────────────────┐
│  Tailscale API  │────▶│  Peer Discovery  │────▶│  LocalSend API  │
│  (mesh network) │     │  (propagator)    │     │  (file xfer)    │
└─────────────────┘     └──────────────────┘     └─────────────────┘

Discovery Flow

  1. Tailscale Status: 
    tailscale status --json
    → get mesh peers
  2. LocalSend Probe: UDP multicast 224.0.0.167:53317 → find localsend-enabled peers
  3. Intersection: Peers on both networks get deterministic Gay.jl colors

Usage

# Discover peers on tailscale with localsend
just ts-localsend-discover

# Send file to peer
just ts-localsend-send <peer> <file>

# Receive mode
just ts-localsend-receive

Python API

from tailscale_localsend import TailscaleLocalSend

tls = TailscaleLocalSend(seed=0x6761795f636f6c6f)

# Discover peers
peers = tls.discover()
# [{'name': 'macbook', 'tailscale_ip': '100.x.x.x', 'localsend_port': 53317, 'color': '#A855F7'}]

# Send file
tls.send(peer='macbook', file='data.json')

# Receive (blocking)
tls.receive(callback=lambda f: print(f"Got {f}"))

Protocol Details

Tailscale Discovery

  • Uses 
    tailscale status --json
    for mesh peers
  • Extracts TailscaleIPs for each peer
  • Falls back to Tailscale API if CLI unavailable

LocalSend Protocol

  • Multicast: 224.0.0.167:53317 (UDP)
  • Announce: JSON with alias, fingerprint, port
  • Transfer: REST API over HTTPS 
    • POST /api/localsend/v2/prepare-upload
    • POST /api/localsend/v2/upload?sessionId=...

Color Assignment

Each peer gets deterministic color from Gay.jl:

peer_color = gay_color_at(hash(peer_fingerprint) % 1000, seed=GAY_SEED)

Integration with epistemic-arbitrage

from epistemic_arbitrage import ArbitrageNetwork

network = ArbitrageNetwork(seed=1069)
for peer in tls.discover():
    network.add_cell(peer['name'], knowledge=peer.get('files', 0))
    
# Propagate knowledge between peers
network.add_propagator(:peer_sync, sources, targets)
network.run_parallel(n_workers=len(peers))

Commands

just ts-peers          # List tailscale peers
just ls-peers          # List localsend peers  
just ts-ls-bridge      # Bridge both networks

Base directory: ~/.codex/skills/tailscale-localsend