Skip to content

MCP Server

Jump to bottom
BLUECARBONS Digital Agency edited this page Jul 3, 2026 · 1 revision

MCP Server

agentic-rss-parser ships a Model Context Protocol (MCP) server that exposes its feed parsing and article fetching capabilities as tools to any MCP-compatible host (Claude Desktop, Cursor, Cline, custom orchestrators, etc.).

Starting the server

# Global install
npm install -g agentic-rss-parser
agentic-rss-mcp

# Local install (from consumer project)
npx agentic-rss-mcp

# Direct from repo clone
pnpm mcp

The server communicates over stdio using the JSON-RPC 2.0 MCP protocol (version 2024-11-05).

Claude Desktop configuration

Add to ~/.config/claude/claude_desktop_config.json (macOS: ~/Library/Application Support/Claude/claude_desktop_config.json):

{
  "mcpServers": {
    "agentic-rss": {
      "command": "npx",
      "args": ["agentic-rss-mcp"],
      "env": {
        "AGENTIC_RSS_MAX_CONCURRENCY": "3"
      }
    }
  }
}

Available tools

fetch_rss_feed

Fetch and agentically analyse an RSS or Atom feed.

Parameters:

Parameter Type Default Description
url string required RSS or Atom feed URL
limit number 10 Max items to return
provider 'heuristic'|'openai'|'anthropic'|'local' 'heuristic' Analysis backend

Returns: JSON object { results, feedErrors } where results is Array<{ item, analysis }>.

Example MCP call:

{
  "name": "fetch_rss_feed",
  "arguments": {
    "url": "https://hnrss.org/frontpage",
    "limit": 5,
    "provider": "heuristic"
  }
}

fetch_full_article

Fetch the full plain-text content of an article URL with HTML stripped. Useful for passing article body as context to a subsequent LLM call.

Parameters:

Parameter Type Description
url string Article URL to fetch

Returns: { url, text } — the URL and plain-text body.

Throttling

The server limits concurrent tool executions. Beyond the limit, requests queue and are processed in order — the server never rejects them with an error.

Setting Default Override
Max concurrent tool calls 3 AGENTIC_RSS_MAX_CONCURRENCY env var
Hard cap 16 Not configurable

limit semantics

limit controls how many new items are stored and returned. Items already in the deduplication store are skipped transparently — they do not count against the limit. This means calling fetch_rss_feed twice with limit=5 on the same feed returns items 1–5 the first time and items 6–10 the second time (assuming at least 10 new items in the feed).

Error handling

Scenario Response
Invalid URL -32602 Invalid params
Unknown provider string -32602 Invalid params
Feed fetch failure feedErrors array in result (not a JSON-RPC error)
LLM API key missing -32603 Internal error with descriptive message
LLM timeout (30 s) -32603 Internal error

Clone this wiki locally