Start free
Guide Teach your agent your knowledge

How to teach Jan your knowledge

Updated 2 min read

The short answer

Teach Jan by giving it reusable assistant instructions, then select that assistant for a chat or project. For your saved reading, KnowMan provides search, reading, and question-answering tools through its MCP server. Jan supports MCP, which lets models connect to external tools and data sources.

On this page8 sections
  1. Create an assistant with clear instructions
  2. Choose an assistant for each chat
  3. Understand Jan’s local storage
  4. Use the hosted KnowMan MCP server
  5. Search and ask about your saved reading
  6. Understand permissions and context limits
  7. Keep instructions and knowledge separate
  8. Questions people ask

Create an assistant with clear instructions

Open Settings > Assistants to create, edit, or delete an assistant. Give each one clear instructions that define its role, tone, level of detail, and output format. Jan’s assistant guide recommends specific instructions with preferences and examples.

Choose an assistant for each chat

In the chat input, select + > Use Assistant and choose the assistant you want. You can also assign an assistant when creating a project, so every conversation in that project uses the same persona and instructions automatically. See Jan’s assistant guide.

Understand Jan’s local storage

Jan stores data locally in JSON format. Its desktop agent workspace includes permanent stores for memories, skills, and saved subagent definitions. Jan Agent uses separate user and project stores, and its memory and skills are not shared with the desktop agent workspace automatically. Jan’s data-folder guide explains these storage boundaries.

Use the hosted KnowMan MCP server

KnowMan offers a hosted MCP server at https://api.knowman.io/mcp. It supports OAuth or an API key sent as Authorization: Bearer <key>. For clients with MCP OAuth, add the server URL, sign in, and approve access. KnowMan’s MCP repository documents the connection flow.

Search and ask about your saved reading

KnowMan provides six MCP tools: kb_search, kb_ask, kb_get_item, kb_save, kb_recent, and kb_memories. Once the server is connected, ask the assistant to search your KnowMan or save a URL. KnowMan’s MCP repository lists the tools and example requests.

Understand permissions and context limits

Jan asks you to approve each MCP tool call unless you allow all tools in Settings. Active MCP connections can use part of the model’s context window and may affect response performance. Jan’s MCP guide also notes that the selected model must support tool calling.

Keep instructions and knowledge separate

Use Jan assistants to set role, tone, detail, and formatting through reusable instructions. Use KnowMan MCP tools for saved articles, links, questions with citations, recent items, and memories. Jan’s assistant guide and KnowMan’s MCP repository describe these separate roles.

Questions people ask

What is the easiest way to teach Jan your knowledge?

Start with an assistant that contains clear, specific instructions. You can choose it from the chat input or assign it to a project. See Jan’s assistant guide.

Can Jan search your saved articles?

Jan supports MCP as a way to connect models with external tools and data sources. Jan’s MCP guide describes that support, while KnowMan’s MCP repository lists the search and reading tools provided by KnowMan.

Does Jan remember files between chats?

Jan has permanent local stores for memories, skills, and saved subagent definitions. General file and shell tools are limited to permitted workspace locations. See Jan’s data-folder guide.

Can KnowMan connect to Jan?

KnowMan provides local and hosted MCP servers, and Jan supports MCP. Use the KnowMan MCP guide for the hosted server URL and supported authentication methods.

Give Jan a knowledge source

Try KnowMan Pro for 14 days with no credit card. Create your account

Start free

Every new account starts with 14 days of Pro. No credit card.