# Orma Memory MCP server

> Provides an episode-first, zero-latency long-term memory layer for local LLMs and AI coding assistants using hybrid SQLite search.

- Repository: https://github.com/wizardwithcodehazard/orma
- Page: https://hostedmcp.org/servers/wizardwithcodehazard/orma
- Transport: stdio
- Stars: 2
- License: Apache-2.0
- Latest version: 0.1.13

## Install

### Claude Code

```sh
claude mcp add orma -- uvx orma-ai
```

### Cursor (`~/.cursor/mcp.json`)

```json
{
  "mcpServers": {
    "orma": {
      "command": "uvx",
      "args": [
        "orma-ai"
      ]
    }
  }
}
```

### VS Code (`.vscode/mcp.json`)

```json
{
  "servers": {
    "orma": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "orma-ai"
      ]
    }
  }
}
```

### Claude Desktop (`claude_desktop_config.json`)

```json
{
  "mcpServers": {
    "orma": {
      "command": "uvx",
      "args": [
        "orma-ai"
      ]
    }
  }
}
```

### Codex (`~/.codex/config.toml`)

```toml
[mcp_servers.orma]
command = "uvx"
args = ["orma-ai"]
```

### Gemini CLI (`~/.gemini/settings.json`)

```json
{
  "mcpServers": {
    "orma": {
      "command": "uvx",
      "args": [
        "orma-ai"
      ]
    }
  }
}
```

## Tools (5)

- `orma_recall`
- `orma_remember`
- `orma_forbid`
- `orma_facts`
- `orma_forget`

## FAQ

### Does Orma require cloud API keys?

No, everything runs locally in SQLite with on-device embeddings, requiring no external servers or cloud API keys.

### How fast is the memory retrieval?

Recall is extremely fast, running in under 25ms using hybrid vector and full-text search without calling an LLM.

### Can I use Orma with existing coding assistants like Cursor or Claude Code?

Yes, it can be used as an MCP server for coding assistants or as an OpenAI-compatible proxy.