hermes-ao

Agent Orchestrator (AO) plugin for Hermes. Gives your Hermes bot 17 tools to spawn, monitor, and manage parallel AI coding agents from chat.

What it does

Once installed, your Hermes bot (Discord, Slack, CLI, etc.) can:

• Spawn agents on GitHub/Linear issues (ao_spawn)
• List active sessions and their status (ao_sessions)
• Send messages to running agents (ao_send)
• Kill sessions (ao_kill)
• Check health of the AO server (ao_doctor)
• Review and verify agent work (ao_review_check, ao_verify)
• Manage issues — create, list backlog (ao_create_issue, ao_backlog)
• Merge PRs created by agents (ao_merge_pr)
• Monitor system observability (ao_observability)
• ...and more (17 tools total)

It also includes a pre_llm_call hook that automatically injects live AO context (active sessions, open issues) into conversations when the user mentions work-related topics.

Prerequisites

• Hermes installed and running
• Agent Orchestrator installed and running somewhere accessible

Install

pip install hermes-ao

Configuration

Set these environment variables before starting Hermes:

Required

Optional

Example setup

# Add to your shell profile (~/.bashrc, ~/.zshrc) or systemd service file
export AO_CWD=/home/user/my-project
export AO_API_URL=http://localhost:3000
export AO_PUBLIC_URL=http://your-server:3000

Then restart Hermes:

hermes restart
# or start a new session

Verify it works

hermes chat -q "What AO sessions are running?"

Or in Discord, ask your bot: "What agents are running?"

Alternative install: directory-based

If you prefer not to use pip, copy the plugin directly:

mkdir -p ~/.hermes/plugins/
cp -r hermes_ao ~/.hermes/plugins/hermes-ao

Set the environment variables and restart Hermes.

Tools reference

Architecture

Discord/Slack/CLI → Hermes → LLM → ao_* tools → AO HTTP API → coding agents

The plugin talks to AO's REST API. If the API is unreachable, it falls back to shelling out to the ao CLI. A circuit breaker prevents hammering a down server.

License

MIT