Mnemosyne Integrations

Mnemosyne runs everywhere. Pick your platform:

Platform	Method	Config
Cursor	MCP (stdio)	`.cursor/mcp.json`
Claude Code	MCP (stdio)	`claude.json`
OpenAI Codex CLI	MCP (stdio)	`.codex/mcp.json`
Windsurf	MCP (stdio)	`.windsurf/mcp_config.json`
OpenWebUI	Native @tool	Workspace tool config
Pi	Pi extension + skill	`pi install npm:@mnemosyne-oss/pi-mnemosyne`
Hermes Agent	MCP + Plugin	`~/.hermes/config.yaml`
Zero	Plugin (tools + hooks)	`.zero/plugins/mnemosyne/`
OpenClaw	Native plugin	`pip install mnemosyne-memory[openclaw]`

Model providers

Mnemosyne's outbound surfaces all speak the OpenAI-compatible protocol, so any provider that implements it works with environment variables alone.

Provider	Surfaces	Config
Any OpenAI-compatible vision endpoint	Vision	`MNEMOSYNE_MODALITY_*`
Atlas Cloud	Chat, embeddings, vision	`MNEMOSYNE_LLM_`, `MNEMOSYNE_EMBEDDING_`, `MNEMOSYNE_MODALITY_*`

Media understanding is off by default — MNEMOSYNE_MODALITY_ENABLED must be set before Mnemosyne makes any outbound call to describe a file.

{
  "mcpServers": {
    "mnemosyne": {
      "command": "mnemosyne",
      "args": ["mcp"],
      "env": {}
    }
  }
}

Make sure the mcp extra is installed:

pip install "mnemosyne-memory[mcp]"

That's it. Three tools become available: mnemosyne_remember, mnemosyne_recall, mnemosyne_forget.

Use the Python API directly:

from mnemosyne import remember, recall
remember("User prefers dark mode")
results = recall("user preferences")