AKADATA LIMITED

AKAMAN — Local System Documentation for AI Agents

AKAMAN is a small C utility that retrieves compact fragments from documentation already installed on the host: man pages, /usr/share/doc, and system headers.

AKAMAN logo

AKAMAN

A small Linux C utility that retrieves the compact documentation fragment an AI agent needs from the host where it is running.

What AKAMAN is, and the problem it solves

A coding agent frequently needs one small piece of authoritative local documentation: the syntax of an option, the grammar of a subcommand, the meaning of an exit status, or the declaration of a C function. That documentation is often already installed on the machine the agent is running on — a manual page, a file under /usr/share/doc, or a system header. The hard part is not finding it; it is avoiding the cost of placing all of it into the model’s context.

Reading documentation locally is not the same as putting all of it into model context.

AKAMAN performs the local discovery and extraction and returns only the useful fragment.

AKAMAN is deliberately narrow. It is not an AI model. It does not summarise documentation with AI. It does not maintain a copied documentation corpus, does not use embeddings, does not use a vector database, does not require a cloud lookup, and does not create a second documentation authority. The locally installed documentation remains the authority; AKAMAN is a precise local reference tool rather than a documentation chatbot.

To use AKAMAN, the agent talks to it through the Model Context Protocol (MCP). MCP is a structured way for an AI application to connect to a server, discover the tools it exposes, and call them with typed arguments. AKAMAN exposes exactly one tool, man, and returns a bounded, authoritative fragment instead of an entire manual. The rest of these pages explain how that works in practice.