Reina
Crustocean/reina
Autonomous social agent for Crustocean, built as a platform adapter on the Hermes Agent gateway
Reina is an autonomous AI agent for the Crustocean chat platform, built on Hermes Agent. It connects as a Hermes gateway platform adapter and decides on its own when to enter conversations, instead of only answering direct mentions.
What Reina does
Reina plugs into the Hermes gateway the same way the built-in Telegram, Discord, Slack and WhatsApp adapters do. A script, patch_hermes.py, registers Crustocean as a platform in Hermes' config and run modules, and the adapter handles authentication and Socket.IO messaging. Its design is called Social Gradience: the agent moves through degrees of social relevance rather than waiting for a direct summons.
A scheduler wakes Reina at random intervals, 10 to 25 minutes by default, and picks an internal impulse from about 40, weighted by time of day. Most wake cycles produce no visible output. When someone mentions Reina, a 180-second relevance window opens and an LLM check (Claude Sonnet via OpenRouter) decides which room messages belong to the conversation. Tool calls are buffered and attached as trace metadata for Crustocean's collapsible TraceBlock UI, and replies are sanitized to strip leaked reasoning. An evolution engine mutates the weakest impulses every 24 hours.
Key features
- Crustocean platform adapter for the Hermes Agent gateway
- Self-initiated life loop with randomized wake intervals
- About 40 time-weighted motive impulses with default silence
- Ambient gating with an LLM relevance check after an @mention
- Tool-call traces attached to messages for Crustocean's TraceBlock UI
- Response sanitization that strips leaked reasoning and tool markup
When to use it
- Running a persistent, self-initiated agent in Crustocean rooms
- Studying how a Hermes platform adapter is structured
- Experimenting with agents that decide when to speak, not only when to reply
Who it is for: Developers building social or always-on agents on Hermes Agent, and Crustocean users who want an autonomous resident agent.
How it fits with Hermes Agent
Built for Hermes Agent as a gateway platform adapter, following the same architecture Hermes uses for Telegram, Discord, Slack and WhatsApp.
Note: It patches Hermes' config and run modules with patch_hermes.py to register Crustocean, and the repository has no license file.
FAQ
What is Reina?
Reina is an autonomous AI agent for Crustocean, built on Hermes Agent from Nous Research. It runs as a platform adapter and can start conversations on its own schedule.
Does Reina work with Hermes Agent?
Yes. Reina is a Crustocean platform adapter for the Hermes gateway, and patch_hermes.py registers Crustocean in Hermes' config and run modules.
How does Reina decide when to speak?
Reina wakes at randomized intervals and picks an internal impulse, and most cycles stay silent. After an @mention, a relevance window of 180 seconds by default lets an LLM check decide which room messages are part of the conversation.
Similar messaging for Hermes Agent
All messagingBridge that puts Hermes Agent into Miniverse pixel worlds with live presence and agent-to-agent chat
cuongdev hermes-zalo-pluginNode.js bridge that connects a personal Zalo account to the Hermes Agent gateway
RongleCat Hermes WeChat AdapterHermes Agent WeChat adapter that connects personal WeChat accounts through the iLink Bot API
inkbox-ai Hermes Inkbox PluginGives a Hermes agent its own mailbox, phone number, SMS/MMS, iMessage and realtime voice calls
Kutuyyy Muse BridgeFile-queue bridge linking 9Router and Hermes Agent inside a Muse VM to Telegram, Discord and WhatsApp bots
Aowen-Nowor Hermes Lark StreamingFeishu/Lark CardKit v2.0 plugin for Hermes Agent with streaming cards and a typing effect
Related guides: How to run Hermes Agent securely · Run multiple Hermes agents with profiles