Preise werden geladen…
Zipp MCP Server

Talk to Zipp from any AI

Multi-language crypto news with sentiment, importance, and source attribution — wired into your favourite MCP client with a single connector.

uvx zipp-mcp
8 languages · 6 tools · no API key
MCP, in 30 seconds

One protocol every AI client already speaks.

The Model Context Protocol is an open standard that lets AI applications connect to external data and tools through the same contract. Claude Desktop, Claude.ai web, ChatGPT, Cursor, Windsurf, Gemini CLI and a growing list of clients support it. Connect Zipp once — it shows up in every chat.

  • Talk to any data, from any AI. One contract, independent of client, model or vendor.
  • No bespoke integration. One protocol, every client — no webhook glue to maintain.
  • Editorial layer included. Sentiment, importance and attribution are values Zipp already produces — MCP ships them as-is.
Why Zipp

Not just news — signal.

Most news APIs hand you raw text. Zipp ships every story with the signals a market actually cares about — so your AI client can reason about the next step, not just summarise the headline.

Signal
What it does
Sentiment
BULLISH NEUTRAL BEARISH on every story
Importance
0–100; ≥75 is the breaking threshold
Multi-language
8 languages, native-quality
Attribution
original publisher on every item
Taxonomy
7 × 5 stable category slugs
The six tools

One tool per Zipp surface.

The AI client picks which tool to call from natural language in the chat. For the full JSON contract and sample responses see the README.

search(query, lang?, category?, limit?)

Full-text search across Zipp's news catalogue. Returns recent matching stories ordered by recency (with relevance as a tiebreaker). Use for questions like "what's happening with Bitcoin ETFs?" or "find news about Solana hacks".

{ "name": "search", "arguments": { "query": "Bitcoin ETF inflows", "lang": "en-US", "limit": 5 } }

get_latest(lang?, category?, limit?)

Latest news from the last 24 hours. Optionally scoped to a category. Returns posts ordered newest-first. Use for "what's new today?" or "what happened in DeFi today?".

{ "name": "get_latest", "arguments": { "lang": "en-US", "category": "ecosystems-ethereum", "limit": 10 } }

get_breaking(lang?, limit?)

Breaking news only — last 24 hours, importance score ≥ 75. Lower volume than get_latest but every item is market-moving. Use for "what's the most important crypto news right now?".

{ "name": "get_breaking", "arguments": { "lang": "en-US", "limit": 10 } }

get_featured(lang?, limit?)

Editor-picked feature stories (is_featured=TRUE). No time window. Use when the user wants curated highlights rather than recency-sorted news.

{ "name": "get_featured", "arguments": { "lang": "en-US", "limit": 5 } }

get_post(slug_or_id, lang?)

Full detail of a single post — title, summary, full body, all categories, hashtags, source attribution. Accepts either a slug (from a previous tool call) or a numeric id.

{ "name": "get_post", "arguments": { "slug_or_id": "solana-network-outage-2026", "lang": "en-US" } }

list_categories(lang?)

List the full Zipp taxonomy (7 main groups × 5 leaves = 35 categories total). Use to discover valid category slugs for the search / get_latest tools.

{ "name": "list_categories", "arguments": { "lang": "en-US" } }
Install

Pick your client.

Most users want the Remote tab — zero install, always up to date. Use Local only when your client is stdio-only (older Claude Desktop releases, custom agents).

1. Claude Desktop

~/Library/Application Support/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "zipp": {
      "type": "http",
      "url": "https://zippfeed.com/mcp/"
    }
  }
}

2. Claude Code (CLI)

claude mcp add --scope user --transport http zipp https://zippfeed.com/mcp/

3. Claude.ai web (Connectors)

Settings → Connectors → Add custom connector → paste the URL → no auth. Once connected, Zipp appears in every chat.

https://zippfeed.com/mcp/

4. ChatGPT (Custom Connectors)

Same flow: Settings → Connectors → Add → paste the URL.

https://zippfeed.com/mcp/

5. Cursor

~/.cursor/mcp.json

{
  "mcpServers": {
    "zipp": { "url": "https://zippfeed.com/mcp/" }
  }
}

6. Windsurf

~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "zipp": { "serverUrl": "https://zippfeed.com/mcp/" }
  }
}

7. Gemini CLI

~/.gemini/settings.json

{
  "mcpServers": {
    "zipp": { "httpUrl": "https://zippfeed.com/mcp/" }
  }
}

8. Anthropic Messages API

The Anthropic SDK accepts remote MCP as a first-class tool — no proxy needed.

client.messages.create(
    model="claude-opus-4-7",
    max_tokens=1024,
    mcp_servers=[{
        "type": "url",
        "url": "https://zippfeed.com/mcp/",
        "name": "zipp",
    }],
    messages=[{"role": "user", "content": "What is breaking in crypto today?"}],
)

9. ChatGPT / OpenAI Responses API

The OpenAI Responses API supports the mcp tool type — drop Zipp straight into your ChatGPT flows.

client.responses.create(
    model="gpt-4.1",
    input="What is breaking in crypto today?",
    tools=[{
        "type": "mcp",
        "server_label": "zipp",
        "server_url": "https://zippfeed.com/mcp/",
    }],
)

1. uvx (zero install)

One line — with uv installed, this fetches the package and runs the stdio server.

uvx zipp-mcp

2. pip

pip install zipp-mcp
zipp-mcp --transport stdio

3. Docker

Multi-arch (amd64 + arm64). Pin with :0.1.0 for reproducibility.

docker run -i --rm ghcr.io/deficlow/zipp-mcp
Discovery

Zipp is published everywhere.

Whichever discovery channel your client prefers, Zipp is there — registry, marketplace or awesome-list.

FAQ

A few questions, answered.

Do I need an API key?
No. The hosted endpoint at https://zippfeed.com/mcp/ is public and rate-limited at the Cloudflare edge per IP — same permission model as our RSS and JSON Feed surfaces.
Can I self-host?
Yes. pip install zipp-mcp for the stdio launcher, or docker run ghcr.io/deficlow/zipp-mcp for the container. Source is MIT-licensed at github.com/deficlow/zipp-mcp.
Which AI clients does it work with?
Anything that speaks MCP — Claude Desktop and the Claude.ai web app via Connectors, ChatGPT via the OpenAI Responses API mcp tool, Cursor, Windsurf, Gemini CLI, and the Anthropic Messages API. See the Install section above for one-click recipes.
What languages are supported?
Eight: English, Turkish, Spanish, Russian, Portuguese, French, German, Italian. Pass any of these as the lang argument on tools that accept it — translations are native-quality, not machine-passthrough.
How fresh is the data?
Source agents poll RSS / Telegram / X continuously; the editorial pipeline turns a raw item into a published post in the order of minutes. get_latest and search always read the live database — there is no stale cache layer.
What does this cost?
Free for end users — the parent site is advertiser-supported. There is no metered billing on MCP usage.
Where is the canonical discovery doc?
At /.well-known/mcp.json. Clients that follow the well-known discovery convention can pick up the server URL and capabilities from there.
Connect now

Zipp in your AI client in five minutes.

One connector — then ask "what's breaking in crypto today?" in any chat. Editorial signals included.