Agent Magnet

Name: Agent Magnet
Author: helinakdogan

by helinakdogancommunityv0.1.11

An MCP server for persistent AI memory that stores and retrieves user profiles, behavioral signals, and session context. It supports tools for injecting memory into prompts, recording signals, getting a learned profile, and generating a cold-start profile for new users.

From the registry: Behavioral memory layer for AI tools. Learns preferences and context across Claude, Cursor, Codex.

GitHub Repository

Install & configure

Prerequisites

required
Install Agent Magnet
Install the Agent Magnet Python package. The README shows both pip and pipx; pipx is the simplest isolated install path for running the MCP server command.
```
$ pipx install agent-magnet
```
required
Provision Redis
Create a Redis instance or obtain a Redis connection URL for the behavioral memory layer. The README suggests Upstash as one option.
https://upstash.com
required
Get an OpenAI API key
Obtain an OpenAI API key used by the reflector model.

Install in your client

Paste into the AI you're chatting with — it will figure out the rest

Please install the `agent-magnet` MCP server into my current AI client (that's you).

Required prerequisites (do these first if not already done):
- **Install Agent Magnet** — Install the Agent Magnet Python package. The README shows both pip and pipx; pipx is the simplest isolated install path for running the MCP server command. Run: `pipx install agent-magnet`
- **Provision Redis** — Create a Redis instance or obtain a Redis connection URL for the behavioral memory layer. The README suggests Upstash as one option. (https://upstash.com)
- **Get an OpenAI API key** — Obtain an OpenAI API key used by the reflector model.

Canonical MCP server config (stdio transport):
- command: `agent-magnet-mcp`
- args: []
- required environment variables:
  - `MAGNET_REDIS_URL`: Redis connection URL for the behavioral memory layer. (example: `<your-redis-url>`)
  - `MAGNET_OPENAI_KEY`: OpenAI API key used by the reflector model. (example: `<your-openai-key>`)
- optional environment variables:
  - `MAGNET_USER_ID`: User identifier used by some clients/setup flows to load and save memory for a specific user. (example: `<your-user-id>`)
  - `MAGNET_PROJECT_ID`: Project identifier used by some clients/setup flows to scope memory. (example: `<your-project-id>`)
  - `QDRANT_URL`: Qdrant URL for the episodic memory layer. (example: `<your-qdrant-url>`)
  - `NEO4J_URL`: Neo4j URL for the knowledge graph layer. (example: `<your-neo4j-url>`)

Note: README provides MCP server setup via stdio using the `agent-magnet-mcp` command. Core MCP example includes `MAGNET_REDIS_URL` and `MAGNET_OPENAI_KEY`; client-specific examples additionally include `MAGNET_USER_ID` and `MAGNET_PROJECT_ID` for automatic memory load/save workflows. The Claude Code Stop hook is client-specific and not part of the canonical MCP server config.

Add this MCP server to my current client's config in the correct format for you. If you need secrets or credentials I haven't provided, ASK me — do not invent values or leave raw placeholders. After adding it, tell me how to verify the server is connected.

Note: README provides MCP server setup via stdio using the `agent-magnet-mcp` command. Core MCP example includes `MAGNET_REDIS_URL` and `MAGNET_OPENAI_KEY`; client-specific examples additionally include `MAGNET_USER_ID` and `MAGNET_PROJECT_ID` for automatic memory load/save workflows. The Claude Code Stop hook is client-specific and not part of the canonical MCP server config.