# Mnemo Cortex MCP server

> Gives AI agents persistent local memory with semantic recall, structured facts, and cross-agent context sharing.

- Repository: https://github.com/GuyMannDude/mnemo-cortex
- Page: https://hostedmcp.org/servers/guymanndude/mnemo-cortex
- Transport: —
- Stars: 160
- License: MIT

## Environment variables

| Name | Required | Secret | Description |
|---|---|---|---|
| `BRAIN_DIR` | no | no |  |
| `MNEMO_AUTO_CAPTURE` | no | no |  |
| `MNEMO_DUMP` | no | no |  |
| `MNEMO_DUMP_DIR` | no | no |  |
| `MNEMO_INSTALL_CONFIG_DIR` | no | no |  |
| `MNEMO_INSTALL_DRY_RUN` | no | no |  |
| `MNEMO_INSTALL_SYSTEMD_DIR` | no | no |  |
| `MNEMO_INSTALL_VENV_DIR` | no | no |  |

## FAQ

### What does Mnemo Cortex do for an AI agent?

It gives agents persistent, local, cross-agent memory that captures what happened and recalls what matters across sessions. It runs on your own machine with no cloud required.

### How does Mnemo Cortex separate real memories from raw conversation logs?

At save time it sorts memories into two tiers: distilled Smart Notes classified into eight categories, and raw Session Logs that are excluded from default recall.

### Can I use Mnemo Cortex with ChatGPT?

ChatGPT does not support local stdio MCP, so Mnemo ships a hardened two-route gate that a Custom GPT calls via Actions while the memory server stays private.

### Which MCP clients work with Mnemo Cortex?

Claude Code, Claude Desktop, and any local MCP client speaking stdio transport, including LM Studio, AnythingLLM, OpenClaw, Agent Zero, and Hermes.

### Does Mnemo Cortex send data to the cloud?

No. It runs locally with no telemetry and no cloud dependency.