Hcom

hcom - multi-agent communication

[](https://pypi.org/project/hcom/) [](https://opensource.org/licenses/MIT) [](https://www.rust-lang.org/) [](https://github.com/aannoo/hcom/stargazers)

Let AI agents message, watch, and spawn each other across terminals

Prefix any agent with `hcom` and they're connected.

Works with Claude Code, Gemini CLI, Codex, OpenCode, and any AI tool that can run shell commands.

https://github.com/user-attachments/assets/1ce23ed9-f529-4be0-8124-816aa4c2fd43

---

Install

brew install aannoo/hcom/hcom

Other install options

# Shell installer for macOS, Linux, Android (Termux), and WSL
curl -fsSL https://github.com/aannoo/hcom/releases/latest/download/hcom-installer.sh | sh

# With PyPI
uv tool install hcom  # or: pip install hcom

---

Quickstart

Run agents with `hcom` in front:

hcom claude  # / gemini / codex / opencode

Prompt:

• send a message to codex
• review what claude did
• spawn 2x gemini and subscribe to their file edits

Open the TUI:

hcom

---

What agents can do

**Spawn**: launch one or many, resume, fork, kill. Any terminal emulator. Scriptable workflows.

**Observe**: read other agent's transcript slices, inspect file edits and commands, view and inject into terminal screens.

**Subscribe**: watch when others are idle, blocked, file changes, specific events. Act on them automatically.

**Message**: direct, broadcast, groups, across devices. Bundle file/transcript/event attachments for handoffs.

---

How it works

hcom acts as a local message bus and event log.

# hooks record activity to db and deliver messages from the db
agent → hooks/cli → db → hooks/pty → other agent

When an agent does something, hooks record it as an event. When other agents do something, hooks deliver relevant events (unread messages, file edit collisions, watch subscriptions) to them from the database. If an agent is idle, pty wakes it so hooks can deliver new events immediately.

At session start, agents get a ~700-token hcom cli primer through hooks. Hooks sit in each tool's config directory and fire automatically with no modifications to Claude, Codex, OpenCode, or Gemini.

---

Multi-agent workflows

Bundled and user scripts (`~/.hcom/scripts/`) for multi-agent patterns:

hcom run                   # list available scripts
hcom run debate "topic"    # run one
hcom run docs              # tell agent to run this to create any new workflow

Included Scripts

**`hcom run confess`** - An agent (or background clone) writes an honesty self-eval. A spawned calibrator reads the target's transcript independently. A judge