Technical Guide to Mana: Automated Knowledge Base Maintenance

Practical Summary and Core Functionality

Mana is an AI-maintained wiki system designed to transform unstructured data—such as research papers, transcripts, and articles—into a structured, linked, and cited knowledge base. Unlike traditional note-taking applications that require manual organization, Mana utilizes an agentic workflow to process files placed in a designated /raw directory. The system automatically extracts durable ideas, updates existing pages, links related concepts, and identifies contradictions within the source material. All operations are performed on plain markdown files, ensuring that the knowledge base remains portable, inspectable, and independent of proprietary database formats. By automating the synthesis of information, Mana allows users to focus on deep work while the agent handles the structural maintenance of the wiki.

Prerequisites and System Architecture

The primary prerequisite for using Mana is a willingness to manage knowledge in a git-backed, markdown-centric environment. Because Mana operates on plain files, it is compatible with standard markdown editors and is specifically optimized for users who prefer Obsidian-style workflows. The architecture relies on git for version control, meaning every agent operation is recorded as a commit. This allows users to review, diff, revert, branch, or clone their entire knowledge base at any time. The system is designed for users who prioritize data ownership and wish to avoid vendor lock-in while leveraging AI for synthesis and organization. The reliance on git ensures that the knowledge base remains portable and inspectable, as there is no proprietary database or lock-in mechanism involved in the storage of the markdown files.

Documented Workflow and Ingestion

The workflow begins by placing source materials, including PDFs, transcripts, and articles, into the /raw directory. Once files are ingested, the Mana agent performs several automated tasks: it reads the content, extracts key concepts, and weaves them into the wiki structure. The agent also adds citations to claims and flags potential contradictions for human review. As the knowledge base grows, users can ask questions, which the agent answers by generating synthesis pages that are automatically linked to relevant concepts. This process ensures that every new source contributes to the existing structure rather than remaining isolated. The agent effectively acts as a curator, turning a pile of disparate documents into a cohesive, living map of learned information.

Reviewing Changes and Maintaining Integrity

A critical component of the Mana workflow is the review process. Because every agent operation is a commit, users can inspect changes via git diffs before finalizing them. This provides a layer of human oversight, allowing users to verify the agent's synthesis, check for accuracy in citations, and resolve the contradictions flagged by the system. The agent also performs linting and reports on the state of the wiki, such as identifying duplicate pages or merging information. This iterative loop ensures that the knowledge base remains accurate and well-organized over time, even as the volume of ingested material increases. By treating the wiki as a git repository, Mana provides a transparent audit trail for every change made by the AI agent.

Limitations and Use Cases

Mana is best suited for individuals who accumulate knowledge over long periods and require a living map of their research. It is particularly effective for deep-dive projects, such as reading books or tracking complex topics across multiple sources. However, users should be aware that the system requires a commitment to a git-based workflow. While it offers an iOS app for mobile capture and reading, the core strength lies in the desktop-based review of git diffs. It is not an all-in-one workspace in the traditional sense, but rather a tool for maintaining a structured, portable, and self-organizing markdown knowledge base. Users should evaluate their comfort with git operations before adopting the tool as their primary knowledge management system. For more information, visit https://getmana.md.

Operational Scope and Data Ownership

The operational scope of Mana extends beyond simple file storage. It actively manages the relationships between concepts by linking pages and creating indexes. Because the system uses plain markdown, users maintain full ownership of their data. There is no proprietary format, meaning the knowledge base can be opened in any standard markdown editor. The agentic nature of the tool means that it does not just store files; it actively improves the knowledge base by merging duplicate pages and flagging contradictions. This proactive maintenance is designed to reduce the cognitive load on the user, allowing them to focus on the content of their research rather than the mechanics of organizing it. The system is built for long-term knowledge accumulation and portability.