Why Use an MCP Server Instead of Raw APIs?

Zero glue code

Modern LLM tool-calling frameworks (OpenAI Agents SDK, LangChain, CrewAI, Anthropic Claude, Google ADK, etc.) natively understand MCP. A one-line adapter call ingests every JazzMCP tool and hands them to the model.

Streaming-first

JazzMCP pushes long-running results (crawler output, exec_py logs) over Server-Sent Events, so your agent can think while data flows.

Schema transparency

Each tool advertises a JSON schema; the LLM knows valid parameter names, types, and constraints—no hallucinated arguments.

Future-proof

Add new endpoints (e.g., summarize_pdf, generate_image) on the server and all clients learn them automatically at the next list_tools refresh.