Metadata-Version: 2.4
Name: tau2-mock-mcp-eval
Version: 0.2.0
Summary: Minimal dependency-free tau2 mock-domain MCP server.
Requires-Python: >=3.9
Description-Content-Type: text/markdown

# tau2-mock-mcp-eval

Minimal dependency-free stdio MCP server for the tau2 `mock/create_task_1`
AgentStudio preprod evaluation path.

This package is intentionally self-contained. It does not import the
official tau2 repository at runtime, because cloud script deployment runs
in Function Compute and cannot see local paths such as
`/private/tmp/tau2-bench-smoke`.

Version `0.2.0` also avoids the Python `mcp` SDK dependency so preprod
script deployment can answer MCP `initialize` within the platform timeout.

## Tools

- `get_users`
- `create_task`
- `update_task_status`
- `transfer_to_human_agents`
- `get_call_log`

## Local Smoke

```bash
python -m pip install -e .
tau2-mock-mcp-eval
```

The command starts a stdio MCP server. Use an MCP inspector/client to
call tools locally.

## Build Wheel

```bash
python -m pip wheel --no-deps -w dist .
```

Then either publish the package to PyPI, or upload the generated wheel to
a URL that Function Compute can download.
