This MCP server provides a Jinja2-based template rendering service exposed via the MCP protocol. It supports rendering templates with complex JSON data, file-based templates, and safe sandboxed execution to prevent misuse. The server implements dual transport modes: standard input/output (stdio) for direct integration with MCP clients, and HTTP-like streamable transport for debugging or testing scenarios. Core tools include render_template for in-memory template rendering, render_template_file for rendering using template files, validate_template to check syntax and usage, list_filters to enumerate Jinja2 filters, and get_template_info to analyze templates and performance characteristics. You can connect clients (such as AI models or automation bots) that supply a template and variables, and receive the rendered output or structured validation results.

Prerequisites:

• Python 3.8+ (recommended 3.12+)
• Git

1. Clone the repository git clone https://github.com/WW-AI-Lab/jinja2-mcp-server.git cd jinja2-mcp-server

2. Create and activate a virtual environment python -m venv .venv

   Linux/macOS

   source .venv/bin/activate

   Windows

   .venv\Scripts\activate

3. Install dependencies pip install -r requirements.txt

4. Run the server (stdio transport for MCP clients) python run_server.py --transport stdio

5. (Optional) Run HTTP/streamable transport for debugging python run_server.py --transport streamable-http --port 8123

Notes and tips:

• The server uses a sandboxed environment for template execution and enforces execution timeout and loop limits to prevent resource abuse.
• You can configure the server further via the JSON configuration options described in the docs (jinja, security, logging, etc.).
• If you plan to expose an HTTP transport, ensure proper security measures are in place (authentication, TLS in production).
• Environment variables like JINJA_AUTOESCAPE, SECURITY_MAX_LOOP_ITERATIONS, and LOGGING_LEVEL can influence behavior; refer to the documentation for recommended values.
• When testing with MCP Inspector, start the HTTP transport and connect to http://localhost:8123 to visualize requests and responses.