typecast-mcp

A validator-only MCP server. It checks a job-posting JSON against a strict Pydantic schema and returns one line per error, in a form an agent can act on:

skills: has 70 items, maximum is 30 - remove 40, starting with required=false ones

No LLM, no API key, no network calls. Dependencies: mcp and pydantic.

It is the validation layer of Typecast, a project that measures how often LLMs fail to return valid structured output and what fixes it.

Tools

• get_job_posting_schema: the JSON Schema of a job posting. It does not carry every rule: duplicate skill names and the intern/experience rule are only checked by the validator.
• validate_job_posting: takes one argument, posting, the whole job-posting JSON as ONE string. Returns valid, a list of error lines (field: problem (got: value)), and the parsed posting when valid.

Install

Requires uv. Add this to your MCP client config (Claude Desktop, Cursor, and others):

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

Claude Code:

claude mcp add typecast -- uvx typecast-mcp

On Windows, if the client cannot find uvx, put the full path to uvx.exe in command. Without uv: pip install typecast-mcp, then use typecast-mcp as the command.

What it does not check

Whether the values are correct. A posting can be valid and still have the wrong seniority or the wrong skills. See Why not just use native structured outputs?

Full spec, limits and design decisions: docs/MCP.md.

License

MIT