# Typecast MCP server

> Converts messy job descriptions into schema-validated, typed JSON using LLM agents and error-feedback retries.

- Repository: https://github.com/nilayp-codes/typecast
- Page: https://hostedmcp.org/servers/nilayp-codes/typecast
- Transport: stdio
- Stars: 1
- License: MIT
- Latest version: 0.1.0

## Install

### Claude Code

```sh
claude mcp add typecast -- uvx typecast-mcp
```

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

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

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

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

### VS Code CLI

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

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

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

### Codex

```sh
codex mcp add typecast -- uvx typecast-mcp
```

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

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

### Gemini CLI

```sh
gemini mcp add -s user typecast uvx -- typecast-mcp
```

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

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

## FAQ

### How does the validation process work?

The server uses a Pydantic model to generate a JSON schema, validates the LLM's raw string output, and performs up to 3 retries with error feedback if validation fails.

### Does using schema mode guarantee correct data?

No, schema mode guarantees the JSON shape and structure, but it does not guarantee the semantic correctness of the content.

### What are the different execution setups available?

There are four setups: A (plain prompt, no retry), B (plain prompt with error-feedback retry), C (provider schema mode, no retry), and D (schema mode with retry).