Skip to content

AgentComms

Bidirectional communication between AI agents and humans via voice calls and chat messaging.

Overview

AgentComms enables AI coding assistants like Claude Code to communicate with humans through:

  • Phone Calls: Real voice calls via Twilio
  • Chat Messaging: Discord, Telegram, WhatsApp via omnichat

Two communication modes:

Mode Direction Use Case
OUTBOUND Agent → Human AI needs input, reports completion, escalates blockers
INBOUND Human → Agent Interrupt agent, send instructions, coordinate tasks

How It Works

                           AgentComms
                    ┌──────────────────────┐
                    │                      │
  ┌──────────┐      │   ┌────────────┐     │      ┌──────────┐
  │ AI Agent │ ────▶│   │ MCP Server │     │◀──── │  Human   │
  │ Claude / │      │   │ (OUTBOUND) │     │      │ (Discord │
  │ Codex    │ ◀────│   └────────────┘     │────▶ │  Phone)  │
  └──────────┘      │                      │      └──────────┘
                    │   ┌────────────┐     │
                    │   │   Daemon   │     │
                    │   │ (INBOUND)  │     │
                    │   └────────────┘     │
                    │         │            │
                    │    ┌────┴────┐       │
                    │    │  tmux   │       │
                    │    │  pane   │       │
                    │    └─────────┘       │
                    └──────────────────────┘

OUTBOUND (MCP Server)

  1. AI needs input → Calls your phone or sends a chat message
  2. You respond → Voice is transcribed, chat is read directly
  3. AI continues → Uses your input to complete the task

INBOUND (Daemon)

  1. You send a message → Type in Discord channel
  2. Daemon receives → Routes to the correct agent via tmux
  3. Agent sees it → Message appears in agent's terminal
  4. Agent polls → Uses check_messages MCP tool to read messages

Features

  • Phone Calls: Real voice calls to your phone via Twilio
  • Chat Messaging: Send messages via Discord, Telegram, or WhatsApp
  • Multi-turn Conversations: Back-and-forth discussions
  • Smart Triggers: Hooks that suggest calling/messaging when stuck
  • Inbound Polling: AI can check for messages during long tasks
  • Event Store: SQLite database tracks all communication

Quick Start

# Build
go build -o agentcomms ./cmd/agentcomms

# Generate configuration
./agentcomms config init

# Run MCP server (for AI → Human)
./agentcomms serve

# Run daemon (for Human → AI)
./agentcomms daemon

See Getting Started for detailed setup instructions.

The plexusone Stack

AgentComms is built with the plexusone voice and chat AI stack:

Package Role
omnivoice Voice abstraction (TTS/STT)
omnichat Chat abstraction (Discord, Telegram, WhatsApp)
omnivoice-twilio Phone calls via Twilio
mcpkit MCP server runtime
elevenlabs-go ElevenLabs TTS/STT
omnivoice-deepgram Deepgram TTS/STT