Helmsman is an MCP server tool crafted in Rust, designed to provide adaptive instructions to various AI coding agents. Developed by Helmsman, this free server specializes in delivering dynamic, context-aware guidance to models like Gemini CLI, Google Antigravity, Claude Code, and Codex. It operates with a local-only architecture, ensuring direct interaction with local AI client binaries. Helmsman tailors instruction verbosity and content based on the AI model's capabilities and the detected operating environment, making it a focused utility for developers seeking precise AI agent interaction.

What does Helmsman connect to?

Helmsman functions as an adaptive instruction server specifically for AI coding agents. This MCP server tool connects with various AI clients, including Gemini CLI, Google Antigravity, Claude Code, and Codex. It acts as an intermediary, delivering tailored instructions to these agents. Built in Rust, Helmsman ensures compatibility across a range of prominent AI coding platforms, enabling developers to manage interactions with their chosen AI models through a unified approach. Its primary role is to serve context-aware guidance to these connected AI agents.

Core features and tools

Helmsman's primary callable tool is instructions, used to deliver dynamic guidance to AI coding agents. This is invoked via "/helmsman:instructions claude-opus-4-5-20251101", enabling precise instruction delivery. The server supports the stdio transport protocol for communication. It features adaptive instruction delivery, adjusting verbosity based on the AI model's capability, such as providing minimal guidance for Opus and more verbose details for Haiku. It also adapts instructions based on the detected operating system, shell, and available tools, and utilizes Jinja2 templates for instruction customization.

Who is it for?

Helmsman is designed for developers and users who actively work with AI coding agents. Its capabilities in adapting instructions based on model ability and environmental factors make it suitable for those needing precise control over AI agent interactions. Users seeking to customize AI agent behavior through Jinja2 templates and dynamic context will find it valuable. Skip this if you require a server with an integrated authentication model or support for SSE or streamable HTTP protocols.

How to get started

No credentials are required for Helmsman itself, as it handles authentication through the underlying AI client. It uses the stdio transport protocol.

Use cases

• Serving dynamic, context-aware instructions to AI coding agents, adapting verbosity based on the model's capability (e.g., minimal guidance for Opus, verbose for Haiku).
• Adapting instructions based on the detected operating system, shell, and available tools in the environment.
• Utilizing Jinja2 templates to customize instructions based on the specific context.

Helmsman offers a specialized solution for managing AI coding agent interactions through adaptive instruction delivery. Operating locally, it excels at tailoring guidance based on AI model capabilities and environmental context. While it relies on underlying clients for authentication, its support for Jinja2 templates provides significant customization. Try invoking the "instructions" tool to deliver context-aware guidance to your preferred AI coding agent.