# prothomalo-mcp

An [MCP](https://modelcontextprotocol.io/) server that lets an AI assistant search and read
[Prothom Alo](https://www.prothomalo.com) — Bangladesh's largest Bengali-language newspaper.

Built with [FastMCP](https://github.com/PrefectHQ/fastmcp). Works in Claude Code, Claude Desktop,
Cursor, or any MCP-compatible client.

> **Unofficial.** This project is not affiliated with, endorsed by, or connected to Prothom Alo
> or Mediastar Ltd. It is an independent client for endpoints the site already serves publicly.

## What it does

| Tool | Purpose |
|---|---|
| `search_news(query, limit)` | Keyword search across the archive. Works in Bengali and English. |
| `latest_news(section, limit)` | Newest articles, optionally filtered to one section. |
| `read_article(url_or_slug)` | Full article text for a single story. |
| `list_sections(contains)` | Browse the ~356 available section slugs. |

Useful for media monitoring, brand and competitor mention tracking, or researching how a topic
is being covered in Bangladeshi press.

## Install

Requires [uv](https://docs.astral.sh/uv/) and Python 3.12+.

```bash
git clone https://github.com/Claudefarid/prothomalo-mcp.git
cd prothomalo-mcp
uv sync
```

No API key, login, or paid subscription is needed.

### Connect to a client

```bash
uv run fastmcp install claude-code server.py:mcp     # Claude Code
uv run fastmcp install claude-desktop server.py:mcp  # Claude Desktop
uv run fastmcp install cursor server.py:mcp          # Cursor
```

Restart the client afterwards. Then just ask it things like *"What has Prothom Alo published
about startups this week?"*

### Verify it works

```bash
uv run python test_server.py
```

## How it works

Prothom Alo runs on the [Quintype](https://www.quintype.com/) CMS, which exposes a public JSON
API under `/api/v1/`. The site's `robots.txt` permits these paths — it disallows only
`/api/auth/`, `/api/comments/get_comments_json`, and `/story/*/element/`. **No HTML scraping is
involved.**

Endpoints used:

| Endpoint | Used for |
|---|---|
| `/api/v1/search?q=` | Keyword search |
| `/api/v1/stories?limit=` | Latest articles |
| `/api/v1/stories-by-slug?slug=` | Full article body |
| `/api/v1/config` | Section list |

### API quirks worth knowing

If you're building against this API yourself, these cost me some time:

- **`/api/v1/advanced-search` does not respect recency.** Querying `section-name=bangladesh`
  returns articles from 2014. It's an archive search, not a news feed.
- **`section=` is silently ignored.** The parameter is `section-name`. Passing `section` returns
  unfiltered results rather than an error — easy to miss.
- **`/api/v1/stories` returns genuinely current articles but has no working section filter.**
  So `latest_news()` fetches the newest 100 and filters client-side. A quiet section may
  therefore return fewer items than requested; the tool reports this rather than failing silently.

## Please be considerate

Prothom Alo's articles are their copyrighted work. This tool is intended for personal research,
monitoring, and summarization — **not for republishing their content**. Requests identify
themselves via User-Agent. Keep your volume reasonable; don't hammer their servers.

## License

MIT — see [LICENSE](LICENSE). The license covers this code only, not any content retrieved
through it.