Hermes + Honcho + GBrain

What Hermes + Honcho + GBrain does

The guide splits agent memory into two jobs. Honcho, from Plastic Labs, remembers the person, the relationship and the conversation history by reasoning over stored interactions in the background. GBrain remembers the notes, sources, links and evidence, so the agent can show where a claim came from. Hermes stays the outer loop that receives messages from CLI, Matrix, Telegram, API clients, cron jobs and webhooks, calls tools and MCP servers, and decides which layer a task needs.

The repository covers running GBrain as a local HTTP MCP service that Hermes can call safely, a conservative bootstrap script that checks before it changes anything, config templates, security notes and operations pages. It includes AGENTS.md and INSTALL_FOR_AGENTS.md so you can hand the setup to an AI agent, and its first half is written for readers who only want the idea before the plumbing.

Key features

• Explains the Honcho and GBrain memory split and when Hermes should use each
• GBrain exposed as a local HTTP MCP service for Hermes
• Conservative bootstrap script that checks before changing anything
• Config templates, security notes and operations pages
• AGENTS.md and INSTALL_FOR_AGENTS.md for agent-assisted setup

When to use it

• Giving a Hermes agent both relationship memory and source-backed notes
• Letting an AI agent perform the integration from INSTALL_FOR_AGENTS.md
• Deciding whether a task needs conversation memory, a document brain, both or neither

Who it is for: Hermes Agent users who want long-term personal memory and source-grounded knowledge in one setup.

How it fits with Hermes Agent

Written for Hermes Agent as the runtime that routes between Honcho and GBrain, with GBrain connected through MCP.

FAQ