Module mcp 

MCP (Model Context Protocol) server — exposes scraping & pipeline tools MCP (Model Context Protocol) server for graph-based scraping.

Exposes stygian-graph scraping and pipeline capabilities as an MCP server over stdin/stdout using the JSON-RPC 2.0 protocol. External tools (LLM agents, IDE plugins) can scrape URLs, query REST/GraphQL APIs, parse feeds, and execute full pipeline DAGs via the standardised MCP interface.

Enabling

[dependencies]
stygian-graph = { version = "*", features = ["mcp"] }

Running the server

Add stygian-graph as a dependency with the mcp feature and call McpGraphServer::run() from your own binary:

use stygian_graph::mcp::McpGraphServer;

#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
    McpGraphServer::run().await
}

Protocol

Implements MCP 2025-11-25 over JSON-RPC 2.0 on stdin/stdout.

MCP Method	Description
initialize	Handshake, return server capabilities
tools/list	List available scraping and pipeline tools
tools/call	Execute a scraping or pipeline tool

Tools

Tool	Key Parameters	Returns
scrape	url, timeout_secs?, proxy_url?, rotate_ua?	data, metadata
scrape_rest	url, method?, auth?, query?, body?, headers?, pagination?, data_path?	data, metadata
scrape_graphql	url, query, variables?, auth?, data_path?	data, metadata
scrape_sitemap	url, max_depth?	data (JSON array of entries), metadata
scrape_rss	url	data (JSON array of items), metadata
pipeline_validate	toml	nodes, services, execution_order, valid
pipeline_run	toml, timeout_secs?	per-node outputs, skipped, errors
charon_*	feature-gated HAR diagnostics and planning inputs	Charon report/policy JSON payloads

Structs

McpGraphServer

MCP server exposing stygian-graph scraping and pipeline tools.