Skip to main content

​
Grok Build + Memanto

Grok Build Connect Grok Build, xAI’s coding agent, to Memanto so it can store decisions, preferences, and corrections and recall them across sessions. Grok Build reads Claude Code marketplaces, plugins, skills, agents, and hooks alongside its own .grok/ directory, so it uses the same Memanto plugin as Claude Code — no separate package.

​
Quick Start

1

Install the plugin

Add the Memanto marketplace and install the plugin:
Grok Build picks up the installed plugin automatically. Or load it from a local checkout:
2

Install and configure the Memanto CLI

Get a free API key at console.moorcheh.ai, or export it yourself with export MOORCHEH_API_KEY="...". You can also run on-prem with no key at all.
3

Create an agent for your project

4

Start Grok Build

​
Features

  • Memory on its own — the memanto skill loads when you state a decision worth keeping or ask what was decided earlier.
  • Always-current MEMORY.md — synced when a session starts, and again before compaction.
  • Readable transcript — memory operations are summarized in plain English.
  • memory-scout subagent — gathers deep background from memory before substantial work.
  • Memory commands — quickstart, remember, recall, answer, capture, forget, upload, conflicts, sync, session, and status.
Once installed, just talk to Grok:
“Remember that we decided to use React for the frontend” “What do you remember about our authentication approach?” “Use memory-scout to get background before we touch the auth module”

​
Architecture

The plugin integrates through lifecycle hooks, skills, a subagent, and commands: The plugin ships no MCP server. Its only network access is the memanto CLI talking to Moorcheh under your own API key. See the Claude Code page for the full command reference.

​
Configuration

Memanto settings live in the CLI, not in the plugin: MOORCHEH_API_KEY is read from the environment or ~/.memanto/.env. For a self-hosted backend, see On-Prem Configuration.

​
Sharing memory with Claude Code

Grok Build and Claude Code both use Memanto’s active agent, so with one agent active they read and write the same memory. A decision stored from Claude Code is available in Grok Build, and vice versa.

​
Separating Grok Build memory

To keep Grok Build’s memory separate, give it its own agent and activate it before launching:
Switch back with memanto agent activate <other-agent>.

​
Per-project memory

Create one agent per project and activate it when you switch projects:
MEMORY.md is written into the project directory, so each repository keeps its own snapshot.

​
Troubleshooting

Check that Grok Build sees the plugin:
Inside the TUI, /plugins opens the extensions modal. If memanto is missing, re-run the install commands above, or start Grok with --plugin-dir pointing at a local checkout.
The hook stays quiet when it can’t sync. Check the CLI directly:
If memanto isn’t found, install it with pip install memanto and make sure it’s on the PATH Grok Build inherits. If there’s no active agent, run memanto agent create <project-name>.
Memory is per agent. Make sure the active agent is the one that stored the memories — memanto agent list shows every agent.

​
Requirements

  • Grok Build
  • Python 3.10+
  • A Moorcheh account and API key — or on-prem with no key

​
Next Steps