flame_mcp_server

Flame MCP Server exposes a structured interface to Flame engine documentation, tutorials, and cached content for AI assistants and MCP clients. It provides two primary capabilities: a searchable documentation index and a guided set of tutorials (Space Shooter, Platformer, Klondike) that are delivered step-by-step. Clients such as Claude Desktop or Amazon Q CLI can query the server to fetch documentation topics, learn through complete tutorials, and rely on the local cache for fast offline access. Typical interactions include searching for Flame API details, requesting a full tutorial, or listing available tutorials. The server is designed to work with an MCP client by exposing endpoints that respond with structured content, enabling seamless integration into conversational workflows and developer assistants.

Prerequisites:

• Dart SDK: Version 3.2.0 or higher
• Internet connection for initial documentation sync
• MCP client (Claude Desktop, Amazon Q CLI, or compatible client)

Installation steps:

1. Clone the repository and navigate to the project directory: git clone <repository-url> cd flame_mcp_server

2. Build and set up the MCP server (this will install Dart dependencies, compile the server, and cache Flame documentation): ./build_clean.sh

3. Run the server (using the prebuilt executable as documented in the Quick Start): ./build_flame_mcp_live # or the path to flame_mcp_live depending on your environment

4. Configure your MCP client to point at the server executable as shown in the Quick Start instructions, using an absolute path to the built flame_mcp_live binary.

Notes:

• If you modify the repository, re-run ./build_clean.sh to refresh the cache and rebuild the server.
• Ensure the flame_docs_cache directory contains the expected 146 Markdown files after a fresh sync.

Tips and common issues:

• Always use the absolute path to the flame_mcp_live executable in MCP client configurations to avoid “Server Not Found” errors.
• If you encounter no search results, regenerate the cache with ./build_clean.sh and verify flame_docs_cache contains content (e.g., find flame_docs_cache -name "*.md" | wc -l should show around 146).
• For rate limits when syncing documentation, use a GitHub token as described in the Configuration section to increase API quota.
• The server’s cache size is roughly 3 MB and the documentation set contains 146 Markdown files.
• Restart your MCP client after starting the Flame MCP Server to ensure the new content is loaded.