Claude Code Hook Comms (HCOM)
A local message bus and event log that enables AI agents to communicate, observe, and spawn each other across terminals.
What is Claude Code Hook Comms (HCOM)?
hcom (hook-comms) is a multi-agent communication system that connects AI coding tools like Claude Code, Gemini CLI, Codex, and OpenCode. It allows these AI agents to coordinate workflows by messaging each other, observing each other's activities, and spawning new instances, all through a unified local interface. It solves the problem of isolated AI agents by enabling them to collaborate, share context, and automate multi-agent patterns.
Developers and researchers using AI coding assistants (Claude Code, Gemini CLI, Codex, OpenCode) who want to orchestrate multiple agents for complex tasks, collaborative debugging, or automated workflows. It's for users who need agents to communicate, share state, and work together without manual intervention.
Developers choose hcom because it provides a lightweight, local-first message bus that requires no modifications to the underlying AI tools, using hooks to automatically record and deliver events. Its unique selling point is the ability to seamlessly connect existing AI tools into a coordinated multi-agent system with features like cross-device relay, event subscriptions, and pre-built workflow scripts.
Overview
Let AI agents message, watch, and spawn each other across terminals. Claude Code, Codex, Antigravity CLI, Cursor CLI, OpenCode, Kilo, Pi, Kimi
Use Cases
Best For
- Orchestrating debates or collaborative problem-solving sessions between multiple AI agents on a shared topic.
- Setting up persistent file watchers (like the 'fatcow' script) where a headless agent monitors file changes and answers queries from other agents.
- Running honesty evaluation workflows (like the 'confess' script) where agents independently assess and judge each other's outputs.
- Coordinating development tasks across multiple AI agents where one agent needs to review another's transcript, file edits, or terminal activity.
- Connecting AI agents across different machines or development environments via MQTT for synchronized remote collaboration.
- Automating multi-agent launch and management scripts for repetitive workflows like spawning, forking, or killing groups of agents.
Not Ideal For
- Teams that rely exclusively on cloud-based AI assistants without local CLI tools
- Users who need a graphical interface for agent management and monitoring
- Environments with strict security policies that forbid modifying AI tool configs or installing hooks
- Projects requiring support for AI tools outside the listed ones (Claude Code, Gemini CLI, Codex, OpenCode)
Pros & Cons
Pros
Hooks automatically install into AI tool config directories (e.g., ~/.claude) without modifying the tools themselves, enabling instant communication as described in the 'How it works' section.
Allows reading other agents' transcripts, inspecting file edits and commands, and viewing or injecting into terminal screens, providing full context for coordination.
Includes pre-built scripts for patterns like debates and honesty evaluations, plus user-defined scripts in ~/.hcom/scripts/, automating complex multi-agent tasks.
MQTT relay connects agents across different machines via 'hcom relay', enabling remote collaboration without central servers.
Cons
Hooks modify AI tool config directories (e.g., ~/.claude), which can conflict with existing setups and require manual removal with 'hcom hooks remove', posing a risk in shared environments.
Full kill/close functionality only works with kitty, wezterm, and tmux, as noted in the Terminal section, restricting users of other emulators like VS Code integrated terminals.
Only supports four specific AI CLI tools (Claude Code, Gemini CLI, Codex, OpenCode), with no built-in extensibility for others, limiting adoption in diverse AI ecosystems.
Requires configuring hooks, terminals, and MQTT relays, with per-agent settings and scripting, making initial setup steep and error-prone for casual users.
Frequently Asked Questions
Related Projects
SuperpowersAn agentic skills framework & software development methodology that works.
Everything Claude CodeThe agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.
Learn Claude CodeBash is all you need - A nano claude code–like 「agent harness」, built from 0 to 1
Claude Scientific SkillsTurn any AI agent into an AI Scientist. The #1 Agent Skills library for science, used by 160,000+ scientists worldwide. 148 ready-to-use skills plus 100+ scientific databases covering biology, chemistry, medicine, and drug discovery. Compatible with Cursor, Claude Code, Codex, Pi, Antigravity, and the open Agent Skills standard.
Found a gem we're missing?
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.