Starlark
byalexmdac
0Provides a secure Starlark code execution environment for large language model clients.
About
The Starlark MCP server offers a safe and hermetic environment for large language models to execute Starlark code, a Python-like language ideal for untrusted code execution. It captures `print()` function output, limited to 16KB, returning it to the LLM client, and extends Starlark with custom builtin functions to enhance functionality. This enables LLMs to perform computational tasks and generate structured outputs in a controlled manner.
Key Features
- Designed as a plugin for MCP-compatible agents
- Augments Starlark with custom builtin functions
- Captures `print()` output (up to 16KB limit)
- 0 GitHub stars
- Executes Starlark code for LLM clients
- Provides a safe and hermetic execution environment
Use Cases
- Integrating with MCP-compatible agents like Claude Code or Gemini
- Enabling LLMs to perform computational tasks or generate complex outputs
- Executing LLM-generated code in a controlled sandbox
Sponsored