Forgetful

v1.0.0Knowledge & Memorystable

Opensource Memory for Agents

ai-agentsai-memoryclaude-codelong-term-memomcp
Share:
265
Stars
0
Downloads
0
Weekly
0/5

What is Forgetful?

Forgetful is a Model Context Protocol (MCP) server that allows AI assistants like Claude, Cursor, and VS Code to opensource memory for agents

Opensource Memory for Agents

This server falls under the Knowledge & Memory category on MCPgee, the world's largest MCP server directory with 33,000+ servers.

Features

  • Opensource Memory for Agents

Use Cases

Give your AI agent persistent long-term memory across sessions.
Build agents with memory that survives every conversation.
Enable continuous learning and context accumulation.
ScottRBK

Maintainer

LicenseMIT
Languagepython
Versionv1.0.0
UpdatedMay 20, 2026
Statushealthy
Maintenanceactive

Works with

ClaudeOpenAIwindowsmacoslinux

Installation

Manual Installation

npx forgetful

Configuration

Configuration Details

Config File

claude_desktop_config.json

Performance

Response Metrics

Response Time< 200ms
ThroughputMedium

Resource Usage

Memory UsageLow
CPU UsageLow

How to Set Up and Use Forgetful

Forgetful is an open-source persistent memory server for AI agents that stores and retrieves memories, entities, documents, code artifacts, and procedural skills across sessions. It uses local semantic embeddings (BAAI/bge-small-en-v1.5) and a knowledge graph to surface relevant context without sending data to the cloud. Developers use it to give Claude and other agents a long-term memory that accumulates across every conversation, enabling continuous learning and context retention.

Prerequisites

  • Python 3.10+ and the uv package manager installed
  • An MCP client such as Claude Desktop or Cursor
  • SQLite (default) or a running PostgreSQL instance for persistence
  • No external API keys required — embeddings run fully locally
1

Install Forgetful via uvx

The fastest way to run Forgetful is with uvx, which installs and executes the PyPI package forgetful-ai without a manual virtual environment.

uvx forgetful-ai
2

Configure environment variables

Create a .env file (or export variables) to tune memory behaviour. Key settings include the token budget for memory retrieval, the embedding model, the number of auto-links created between related memories, and the backend database.

MEMORY_TOKEN_BUDGET=8000
EMBEDDING_MODEL=BAAI/bge-small-en-v1.5
MEMORY_NUM_AUTO_LINK=3
SERVER_PORT=8020
MAX_GRAPH_LIMIT=2000
DATABASE=sqlite
SQLITE_PATH=./forgetful.db
3

Add Forgetful to your MCP client config

Open claude_desktop_config.json (or your client's equivalent) and add a server entry. Forgetful exposes an HTTP SSE transport on the configured port.

{
  "mcpServers": {
    "forgetful": {
      "command": "uvx",
      "args": ["forgetful-ai"],
      "env": {
        "SERVER_PORT": "8020",
        "DATABASE": "sqlite",
        "SQLITE_PATH": "./forgetful.db"
      }
    }
  }
}
4

Verify the server starts

Start the server manually once to confirm it loads the embedding model and opens the database without errors. The console should print the port it is listening on.

uvx forgetful-ai
5

Explore tool categories in your AI client

Forgetful exposes 42 tools across seven categories: Memory (create, query, update, link), Project, Entity (people and organizations), Code Artifact, Document, Skill, and User. The meta-tool pattern means you will see three primary tool entries that dispatch to the full set.

Forgetful Examples

Client configuration

Minimal claude_desktop_config.json entry using uvx with SQLite storage.

{
  "mcpServers": {
    "forgetful": {
      "command": "uvx",
      "args": ["forgetful-ai"],
      "env": {
        "SERVER_PORT": "8020",
        "MEMORY_TOKEN_BUDGET": "8000",
        "DATABASE": "sqlite",
        "SQLITE_PATH": "/Users/you/.forgetful/db.sqlite"
      }
    }
  }
}

Prompts to try

Example prompts that exercise Forgetful's memory, entity, and skill tools.

- "Remember that my preferred coding style is functional, no classes unless necessary."
- "What do you remember about the architecture decisions we made for the payments service?"
- "Save a skill called 'deploy-prod' with these steps: run tests, build Docker image, push to registry, apply Helm chart."
- "Create an entity for Alice Johnson, senior engineer at Acme Corp, and link her to the payments project."
- "Show me all memories related to our database schema from the last month."

Troubleshooting Forgetful

Embedding model download is slow or fails on first start

Forgetful downloads BAAI/bge-small-en-v1.5 from HuggingFace on first run. Ensure you have internet access during initial startup. Subsequent starts use the cached model. Set HF_HOME to a writable directory if the default cache path causes permission errors.

Server exits immediately with a database error

Check that SQLITE_PATH points to a directory that exists and is writable, or that POSTGRES_* credentials are correct if using PostgreSQL. Run `uvx forgetful-ai` in the terminal first to see the full error before relying on the MCP client log.

MCP client shows no Forgetful tools after adding config

Confirm SERVER_PORT is not blocked by a firewall and that no other process is using the same port. Restart Claude Desktop fully (not just reload) after editing claude_desktop_config.json.

Frequently Asked Questions about Forgetful

What is Forgetful?

Forgetful is a Model Context Protocol (MCP) server that opensource memory for agents It connects AI assistants to external tools and data sources through a standardized interface.

How do I install Forgetful?

Follow the installation instructions on the Forgetful GitHub repository. Clone the repo, install dependencies, and add the server config to your AI client.

Which AI clients work with Forgetful?

Forgetful works with all major MCP-compatible AI clients including Claude Desktop, Claude Code, Cursor, VS Code (GitHub Copilot), Windsurf, and Cline.

Is Forgetful free to use?

Yes, Forgetful is open source and available under the MIT license. You can use it freely in both personal and commercial projects.

Browse More Knowledge & Memory MCP Servers

Explore all knowledge & memory servers available in the MCPgee directory. Each server includes setup guides for Claude, Cursor, and VS Code.

Quick Config Preview

{ "mcpServers": { "forgetful": { "command": "npx", "args": ["-y", "forgetful"] } } }

Add this to your claude_desktop_config.json or .cursor/mcp.json

Read the full setup guide →

Ready to use Forgetful?

Browse our complete directory of 33,000+ MCP servers, read setup guides for your editor, and start building with the Model Context Protocol.

33,000+ ServersFree & Open SourceStep-by-Step Guides