# DTing MCP

This is a small MCP server for DTing. MCP means Model Context Protocol: it lets an
AI client call your own business functions instead of guessing from memory.

## DTing Tools

The server now includes learning-friendly DTing tools in `app/tools/dting.py`:

- `get_dting_company_brief`: returns product positioning and core feature context.
- `suggest_dting_mcp_tools`: explains what kind of MCP tools fit a business goal.
- `draft_dting_support_reply`: drafts a customer support first response.
- `generate_dting_profile_prompt`: creates profile-writing prompts for users.
- `analyze_dting_match_context`: gives a simple explainable match-context score.
- `build_dting_safety_checklist`: returns a trust and safety review checklist.

Think of every MCP tool as:

```text
AI asks a question -> MCP calls your function -> AI gets trusted company data
```

To add your next company tool:

1. Create a normal Python function with clear inputs and output.
2. Register it with `@mcp.tool()`.
3. Keep the docstring simple because AI clients use it to understand the tool.
4. Add a focused test for the pure helper function.

## Docker

Build and start the MCP server:

```bash
docker compose up --build
```

The default URL is:

```text
http://localhost:8004
```

If that port is busy, run it with another port:

```bash
HOST_PORT=8010 PORT=8010 docker compose up --build
```

Stop the server:

```bash
docker compose down
```

## Local Python

Install dependencies and run:

```bash
pip install -r requirements.txt
PORT=8004 python -m app.main
```