Start with the information the system can use
An AI knowledge base gives a system a defined collection of material to consult. That collection could contain articles, notes, reference documents, or recorded decisions. Its usefulness depends on what it contains and whether the relevant information can be found for the task.
For personal research, begin with a question: which information should an answer draw from? You may want evidence from articles you selected, your own notes, or the reasons behind an earlier decision. Identifying those sources makes the scope concrete.
KnowMan builds a knowledge base from your saved reading. It reads articles in full, organizes them automatically, and keeps them with their sources. You can search the saved material or ask questions whose answers cite the reading behind them.
Understand AI knowledge base vs RAG
The terms describe different parts of a system. A knowledge base is the information available for use. Retrieval-augmented generation describes a process: retrieve relevant material, then use that material as context when generating a response. One describes the collection; the other describes a method.
This distinction helps you ask better questions when evaluating a product. Ask what material it can consult, how you can find the original sources, and what happens when your collection does not contain the evidence you need. Those questions concern behavior you can inspect.
KnowMan's verified behavior includes keyword search, Pro semantic search, and answers with citations from saved reading. Those features do not establish its internal retrieval architecture. Evaluate the observable workflow without assuming a particular implementation from the feature names.
Build a collection that supports a real task
Choose a task before deciding what to save. You might be comparing technical approaches, preparing an article, or reviewing the reasoning behind a product choice. Select sources that can contribute evidence to that task, and keep a question ready to test against them.
KnowMan supports saving pages through its Chrome extension or web app. Browser capture and library import are available on Free. Articles organize themselves, and you can add highlights and notes to keep your own observations alongside the saved text.
Pro adds ongoing capture through RSS feeds and newsletters. Rules filter incoming feed and newsletter content. These features help you maintain a collection around sources you follow. They are separate from deciding whether an individual source provides evidence you should trust.
Distinguish finding material from answering a question
Search and Ask serve related purposes. Search helps you locate material to inspect. Ask produces an answer drawn from your reading. Decide which you need at each step: a specific passage, a set of relevant articles, or an explanation supported by sources.
KnowMan includes keyword search on Free, covering saved text, highlights, and notes. Pro adds semantic search for finding material by meaning. Try both with a source you know: first use a phrase from it, then describe its idea using different words.
Ask returns answers with citations from your knowledge base. Free allows five Ask questions daily, while Pro removes that daily limit. Open the cited passages and check whether they support the answer. Citations make that inspection possible; checking the evidence remains part of your work.
Separate source claims from decisions and playbooks
An article may describe an approach worth considering without establishing that you should adopt it. Keep that difference clear in your knowledge base. A source records what someone wrote; a decision records what you chose after considering the evidence and your circumstances.
KnowMan's Memories keep your decisions and their reasons next to the reading behind them. A useful decision explains the choice, the evidence that influenced it, and the conditions that would justify another look. This gives future work a concrete starting point.
Playbooks capture how you want recurring work handled. Connected agents can read your decisions and playbooks through MCP. They can use the same saved context when carrying out a task. Agents cannot write Memories, so reading that context does not grant permission to change it.
Check how agents access and reuse your knowledge
KnowMan connects with Claude Desktop, Claude Code, Cursor, Codex and any agent that supports MCP. In Settings, create an agent key with the permissions you choose, then use the supplied configuration for your client. AI assistant connection is included on Free.
The MCP server has six tools: kb_search searches saved material; kb_get_item reads an item; kb_ask returns a cited answer; kb_memories reads Memories; kb_recent shows recent saves; and kb_save saves a link. Choose a task where those actions have a clear purpose, such as locating references for a draft.
Every new account starts with 14 days of Pro, without a credit card. Use the trial to test your own sources, questions, and agent workflow. Afterward, Free keeps saved articles, keyword search, and five Ask questions daily. Semantic search requires Pro.
- Define which sources should inform the task.
- Inspect cited passages before relying on an answer.
- Record your conclusion separately from source claims.