platform_react_runner

To test React Agent prompts outside an IDE, you use a self-contained LangGraph runner that executes tests on external platforms using configurable LLM and MCP settings. This approach validates prompts in environments closer to production before deployment.

The {{AgentName}} placeholder mechanism works by automatically replacing the tag in a configuration YAML with the contents of a corresponding prompt.md file before execution. This allows dynamic prompt injection without hardcoding agent logic directly into the config.

Yes, you can use MCP tooling and LangGraph to run offline React Agent tests. The runner supports offline operation with zero internal project dependencies, requiring only external libraries like langchain-core and pydantic to function.

The best way to validate agent prompts across different LLM models is by using a platform runner with configurable LLM settings. This allows you to execute the same LangGraph ReAct agent against multiple models to compare behavior in production-like environments.

No, the external platform React runner does not require internal project dependencies. It is fully self-contained, relying only on standard external packages like langgraph, mcp, and pyyaml to execute the prompt testing scripts.

You should choose an external platform runner over an IDE Sub Agent when you need to test prompts using specific MCP tooling and LLM combinations that closely mimic your production environment, which IDE sub agents often cannot fully replicate.