Memory Hive: memory tool for Hermes Agent

File-backed shared memory for AI agent teams with private silos, a shared hive and a curator workflow

What Memory Hive does

Memory Hive is a local, file-backed memory layer that gives each AI agent a private silo and a shared hive. It needs no daemon and no account. The installer creates ~/.memory-hive, sets up the shared hive and adds lightweight boot blocks to the agent platforms it finds on your machine. The README lists Claude Code, Cursor, Codex, Windsurf, Devin, Hermes, aider, Continue and Goose as supported.

Work follows a loop of read, work, write and curate. The memory-hive do command starts a task by optimizing the prompt, recalling relevant memory, matching skills and building a plan. The memory-hive done command logs the result and can record a learning, and a main curator promotes verified lessons into shared knowledge. Skills and memory are Markdown files, the index is SQLite, and the transport is the filesystem. A cohort command emits a plan with 25 specialist roles for the host IDE to fan out to subagents.

Key features

Private per-agent silos plus a shared hive of knowledge and distilled learnings
Three main commands: do, done and status
Curator step that promotes verified lessons into shared knowledge
Markdown memory and skills with an auto-built SQLite index
Cohort command that outputs a 25-role plan for subagent fan-out
Installer that adds one-line boot blocks to detected agent platforms

When to use it

Let several agents, Hermes included, share lessons learned across tasks
Log what an agent did after each job and promote reusable learnings
Split a large task into a 25-role plan and run it with your IDE's subagents

Who it is for: People running more than one AI agent or IDE who want a shared, file-based memory instead of a hosted service.

How it fits with Hermes Agent

Memory Hive is a general memory layer that lists Hermes among the many agents it works with, and its topics include hermes-agent.

How to install Memory Hive

These commands are copied from the project's README. Check the repository for the latest steps before you run them.

curl -fsSL https://memoryhive.neural-forge.io/install.sh | sh

Requirements: POSIX shell and Python 3

FAQ

What is Memory Hive?

Memory Hive is a local memory layer that gives each AI agent a private silo and a shared hive of knowledge. It stores everything as files on your machine and needs no daemon or account.

Does Memory Hive work with Hermes Agent?

The README lists Hermes among the agents Memory Hive works with, and says it works with anything that can read files or run a shell command. The CLI returns plain text and JSON, so any model can call it.

How do I install Memory Hive?

Run the one-line installer script with curl, which creates ~/.memory-hive and wires boot blocks into the agent platforms it detects. Then add an agent with memory-hive add and start work with memory-hive do.