# pi-code MCP server

> Extends the pi coding agent with Claude Code features including checkpoints, memory, web search, subagents, and rule loading.

- Repository: https://github.com/ilovepixelart/pi-code
- Page: https://hostedmcp.org/servers/ilovepixelart/pi-code
- Transport: stdio
- Stars: 36
- License: MIT

## Install

### Claude Code

```sh
claude mcp add pi-code -- npx -y pi-code
```

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

```json
{
  "mcpServers": {
    "pi-code": {
      "command": "npx",
      "args": [
        "-y",
        "pi-code"
      ]
    }
  }
}
```

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

```json
{
  "servers": {
    "pi-code": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "pi-code"
      ]
    }
  }
}
```

### VS Code CLI

```sh
code --add-mcp '{"name":"pi-code","type":"stdio","command":"npx","args":["-y","pi-code"]}'
```

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

```json
{
  "mcpServers": {
    "pi-code": {
      "command": "npx",
      "args": [
        "-y",
        "pi-code"
      ]
    }
  }
}
```

### Codex

```sh
codex mcp add pi-code -- npx -y pi-code
```

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

```toml
[mcp_servers.pi-code]
command = "npx"
args = ["-y", "pi-code"]
```

### Gemini CLI

```sh
gemini mcp add -s user pi-code npx -- -y pi-code
```

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

```json
{
  "mcpServers": {
    "pi-code": {
      "command": "npx",
      "args": [
        "-y",
        "pi-code"
      ]
    }
  }
}
```

## FAQ

### What features does pi-code add to the pi agent?

It adds a todo overlay, checkpoints, memory, web search, subagents, and goals to the pi coding agent.

### How does pi-code handle security and untrusted repositories?

Project resources like MCP servers and hooks are treated as untrusted and only load after user approval. In headless runs, undecided projects load nothing.

### Can I use my existing Claude configuration with pi-code?

Yes, pi-code reads your existing .claude config including rules, commands, skills, hooks, output styles, MCP servers, and agents.