Signal — Privacy-First Chat

Imagine sending a postcard versus a sealed letter. Most messaging apps are postcards — anyone handling them can peek. Signal is the sealed letter, and it’s the only mainstream messenger that treats privacy as the default, not a setting.

Why Signal for AI Agents?

When you let an AI agent send messages on your behalf, you’re trusting it with sensitive conversations. Signal’s end-to-end encryption means only you and the recipient read the messages — not the server, not the carrier, not even Signal itself. For security-conscious workflows, this is a game-changer.

How Hermes Connects to Signal

Hermes uses signal-cli, a Java-based client that runs as a daemon in HTTP mode. The magic happens in two parts:

• Receiving messages: Hermes streams them in real-time via SSE (Server-Sent Events) — think of it as a live news feed for your chats.
• Sending replies: Hermes uses JSON-RPC, a simple remote procedure call, to send responses back.

The best part? No new Python dependencies. Hermes already uses httpx for everything, so you only need to install signal-cli externally.

Setting Up signal-cli

First, install the prerequisites:

# Install Java 17+ (required by signal-cli)
sudo apt install openjdk-17-jre

# Download signal-cli
wget https://github.com/AsamK/signal-cli/releases/latest/download/signal-cli.zip
unzip signal-cli.zip
sudo ln -s $(pwd)/signal-cli/bin/signal-cli /usr/local/bin/

Next, link your phone number as a secondary device:

# Register your number
signal-cli register +15551234567

# Link to your existing Signal app (scan the QR code)
signal-cli link -n "Hermes Agent"

Finally, start the HTTP daemon:

signal-cli daemon --http localhost:8080

Configuring Hermes

In your Hermes agent config, add the Signal adapter:

channels:
  signal:
    enabled: true
    url: "http://localhost:8080"
    phone: "+15551234567"

That’s it. Hermes will now listen for incoming Signal messages and respond automatically.

Practical Tip

Always run signal-cli in a separate, isolated environment (like a Docker container) with firewall rules that only allow localhost access. This prevents any external process from tampering with your Signal daemon — keeping your agent’s communications as private as the protocol itself.

Summary

Signal gives your AI agent a privacy-first communication channel with zero extra Python overhead. By pairing Hermes with signal-cli, you get real-time, encrypted messaging that’s ideal for security-sensitive automation. Start small: have your agent send you a daily digest of your calendar via Signal. You’ll sleep better knowing your digital conversations stay sealed.