Ember

Provides local, persistent memory for LLMs with drift detection and source-linked memories to reduce hallucinations.

other

9

GitHub Stars

other

Language

4 months ago

First Indexed

2 months ago

Catalog Refreshed

Documentation & install

Readme and setup notes from the catalogue, plus a client-ready config you can copy for your MCP host.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "arkya-ai-ember-mcp": {
      "command": "ember-mcp",
      "args": [
        "run"
      ]
    }
  }
}

Ember MCP provides local-first memory for large language models, enabling persistent, context-aware reasoning across sessions and clients while actively managing knowledge freshness to reduce hallucinations. It runs entirely locally, preserving privacy and allowing your AI to reference up-to-date decisions and sources without relying on cloud vector stores.

How to use

You integrate Ember MCP with your MCP clients to give your AI a long-term memory that travels across conversations and tools. Ember continuously evaluates the recency and usage of stored memories, penalizing stale information and highlighting the most relevant, current context. When you switch from one MCP client to another, the AI can pick up where you left off without requiring you to re-summarize everything. Use Ember to keep your architecture decisions, design notes, and discussion outcomes coherent across sessions.

How to install

Prerequisites: ensure you have a supported operating system (macOS, Linux, or Windows with WSL) and Python 3.10 or newer installed on your machine.

Download or prepare Ember MCP using the installer. The installer automatically detects your MCP clients and registers Ember with each one, creates the local storage directory, and downloads the embedding model.
Restart your MCP client applications after installation so Ember can establish the memory bridge with each client.
If you want to verify the setup, run the status command to see which clients are registered and how many memories are stored.

Additional setup and configuration

Manual configuration is possible for MCP clients that aren’t auto-detected. Use the following configuration snippet to register Ember as an MCP server for your client.

{
  "mcpServers": {
    "ember": {
      "command": "ember-mcp",
      "args": ["run"]
    }
  }
}

Available tools

ember_store

Save a named memory with importance level and optional tags

ember_recall

Semantic search with temporal scoring across all memories

ember_deep_recall

Recall plus automatically read source files behind the embers

ember_learn

Auto-capture key information from conversation (facts, preferences, decisions)

ember_contradict

Mark outdated memory stale and store corrected version

ember_list

List all stored memories, optionally filtered by tag

ember_delete

Remove a memory by ID

ember_inspect

View Voronoi cell distribution, statistics, and density

ember_auto

Auto-retrieve relevant context at conversation start with temporal ranking

ember_save_session

Save session summary, decisions, and next steps with source linking

ember_drift_check

Run drift detection — flag stale memories in shifting knowledge regions