Coming from another harness
If you already use Claude Code, Hermes Agent, Cursor or Prime, a lot of what you know carries over. This page maps your habits onto Rhizome, and shows what’s different: memory lives in a folder you can read, and you can see where a run went.
The one big difference
Section titled “The one big difference”In most harnesses, memory is a file or store that belongs to the tool: MEMORY.md, USER.md, a database, a cloud account. In Rhizome, memory is your vault, a folder of Markdown notes. You choose what’s saved with Save to vault. With Prime, answers that use a note say it came from your vault. With Git, every change can be diffed and reverted. See Your vault is the agent’s memory.
From Claude Code
Section titled “From Claude Code”| In Claude Code | In Rhizome |
|---|---|
| Your past sessions | Settings → AI Agents → Import chat history brings them into the vault as notes under Imports/claude-code/. See Import Claude Code history |
CLAUDE.md | The vault’s AGENTS.md. Rhizome keeps a CLAUDE.md alongside it for compatibility |
| MCP servers | Rhizome is an MCP server too. Set Up External AI Tools… (⌘K) adds it to Claude Code’s config, so Claude Code can read and write your vault from the terminal |
| Claude models | Not on Rhizome’s own engine yet; use another provider or Taproot |
| Permission prompts | Permission modes, Limited tools and Power User, set per vault |
/resume | The sessions list. See Sessions |
Rhizome doesn’t run Claude Code inside Chat. Keep using Claude Code in the terminal, connected to your vault through MCP.
From Hermes Agent
Section titled “From Hermes Agent”| In Hermes | In Rhizome |
|---|---|
MEMORY.md / USER.md | Vault notes you save on purpose. Automatic saving exists but is off by default |
| Memory writes you approve | Save to vault is an explicit click. It writes straight away, so review the note after, or undo with Git |
| Observable, interruptible runs | Stop any run; with Prime, send a follow-up while it works; see a Prime run’s file footprint in Mycelium |
| Skills | Rhizome’s own plugin system is planned |
| Your choice of provider | Your own providers, local models or Taproot. Keys live in the system keychain |
From Cursor
Section titled “From Cursor”| In Cursor | In Rhizome |
|---|---|
| Agent chat beside your files | Chat with notes beside it (⌘2 shows Notes) |
| Cursor’s MCP settings | Set Up External AI Tools… (⌘K) writes Rhizome’s MCP entry into ~/.cursor/mcp.json, so Cursor’s agent can use your vault |
| Rules files | The vault’s AGENTS.md |
| Model picker | Your providers or Taproot |
Rhizome isn’t a code editor. Keep Cursor for code, and use Rhizome as the place where what you and your agents learn is kept, reviewed and found again.
Bring your own model
Section titled “Bring your own model”Add a provider or use Taproot. Rhizome’s engine can read, search, edit, and create notes in the open vault. Power User can run one approved shell command in the vault folder. See Set up models and keys.
Plugins
Section titled “Plugins”Rhizome’s own plugin system is planned, not shipped. Other apps can reach your vault through MCP.