mcp-server-llmling

What is mcp-server-llmling?

mcp-server-llmling is a YAML-based configuration system for LLM applications built on the Machine Chat Protocol (MCP). It allows you to define resources, prompts, and tools in YAML files without writing code, providing a complete environment for LLM interaction.

How to use mcp-server-llmling?

Install via uvx mcp-server-llmling@latest and start with a YAML config file using mcp-server-llmling start path/to/config.yml. Configure as a context server in Zed or Claude Desktop by specifying the command and arguments in their respective settings files. Programmatic usage is also supported via Python’s LLMLingServer class.

Key features of mcp-server-llmling

• Static YAML declaration – no code required
• Resources: files, text, CLI output, images, Python callables
• Tools: register Python functions or OpenAPI specs
• Prompts: static templates or dynamic Python functions
• Multiple transports: stdio, SSE, Streamable HTTP

Use cases of mcp-server-llmling

• Define an LLM’s environment and tools in a single YAML file
• Serve content and actions to MCP‑compatible clients (Zed, Claude Desktop)
• Enable hot‑reload for resource files during development
• Expose OpenAPI endpoints as callable LLM tools
• Create reusable tool sets and prompt templates across projects

FAQ from mcp-server-llmling

What MCP protocol operations does it support?

Resource operations (list, read, watch), tool operations (list, execute, get schemas), prompt operations (list, get formatted, completions), and notifications for changes and progress.

What transport options are available?

Default stdio; also SSE and Streamable HTTP for web clients, configurable via transport_options in code.

How do I start the server manually?

Run uvx mcp-server-llmling@latest from command line; or with a config file using mcp-server-llmling start path/to/config.yml.

Can I use it without writing Python code?

Yes – the YAML configuration declares resources, tools, and prompts without requiring custom Python code.

How do I configure OpenAPI‑based tools?

Define a toolset with type: openapi and spec pointing to the OpenAPI schema; Redocly CLI is used automatically if installed for schema bundling.