# Semantic Cache MCP server

> Reduces token costs for AI coding agents by providing a local file cache that returns diffs or minimal updates instead of full file re-reads.

- Repository: https://github.com/CoderDayton/semantic-cache-mcp
- Page: https://hostedmcp.org/servers/coderdayton/semantic-cache-mcp
- Transport: stdio
- Stars: 3
- License: MIT
- Latest version: 0.7.0

## Install

### Claude Code

```sh
claude mcp add semantic-cache-mcp -- uvx semantic-cache-mcp
```

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

```json
{
  "mcpServers": {
    "semantic-cache-mcp": {
      "command": "uvx",
      "args": [
        "semantic-cache-mcp"
      ]
    }
  }
}
```

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

```json
{
  "servers": {
    "semantic-cache-mcp": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "semantic-cache-mcp"
      ]
    }
  }
}
```

### VS Code CLI

```sh
code --add-mcp '{"name":"semantic-cache-mcp","type":"stdio","command":"uvx","args":["semantic-cache-mcp"]}'
```

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

```json
{
  "mcpServers": {
    "semantic-cache-mcp": {
      "command": "uvx",
      "args": [
        "semantic-cache-mcp"
      ]
    }
  }
}
```

### Codex

```sh
codex mcp add semantic-cache-mcp -- uvx semantic-cache-mcp
```

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

```toml
[mcp_servers.semantic-cache-mcp]
command = "uvx"
args = ["semantic-cache-mcp"]
```

### Gemini CLI

```sh
gemini mcp add -s user semantic-cache-mcp uvx -- semantic-cache-mcp
```

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

```json
{
  "mcpServers": {
    "semantic-cache-mcp": {
      "command": "uvx",
      "args": [
        "semantic-cache-mcp"
      ]
    }
  }
}
```

## FAQ

### Is the search semantic using embeddings?

No, the search functionality uses BM25 keyword ranking and grep for exact matches, requiring no model downloads or GPUs.

### Does this server send my code to an external service?

No, everything stays local using a SQLite file. The only network request is a one-time download of the o200k tokenizer file.

### Does it work on Windows?

Yes, the server is compatible with Windows.
