User guide overview

Glean gives you a conversational interface to your company's knowledge. Ask questions, create documents and presentations, analyze data, and get work done, all grounded in what your organization knows. Every response is personalized to you based on your permissions, activity, and role.

For details on how Glean selects and combines knowledge sources, see How Glean accesses information.

What you can do

Start with a question, then build on the answer in the same chat. Use the chat to draft, combine your output with other sources, or hand off the follow-up work. You can use Glean to:

• Ask and find: Get answers with citations from your company's documents, messages, and data. When you type a keyword-style query, search results appear inline so you can jump straight to the right document.

• Create content: Draft, refine, and export documents from what you found using these tools:

  • Canvas — draft and refine documents
  • Slide generation — export documents as slides
  • Spreadsheet generation — export documents as spreadsheets
  • Inline artifacts — show results as a chart or diagram in the response
  • HTML artifacts — export documents as interactive artifacts
  • Podcast artifacts — create podcasts
  • Image generation — generate images from natural-language prompts

• Research: Generate detailed, citation-rich reports that synthesize information from internal systems and the web. See Deep Research.

• Hand off repeat work: Turn a lookup or task you repeat into a Skill that Glean can run for you next time. See Creating a Skill via chat.

• Work with code: Search across repositories and generate code with draft pull requests grounded in your organization's codebase.

• Query data: Ask natural-language questions against Databricks and Snowflake datasets without writing SQL.

• Capture meetings: Record transcripts, summaries, decisions, and action items without adding a bot to the call. See Meeting notes.

• Talk hands-free: Have real-time voice conversations on web, desktop, and mobile.

• Organize and share: Keep your work organized and share it with teammates:

  • Manage chats — rename, pin, and queue chats
  • Projects — group related chats and files
  • Left navigation — customize what appears in the left navigation
  • Chat history — search your past chats
  • Chat sharing — share conversations with teammates

• Resume long-running work: Ask Glean to search your past chats from a new conversation.

Which model should I use?

You choose a model from the filter icon in the chat input. The models in the list depend on what your admin has turned on.

• Keep Auto selected. Auto is the default, and it handles most questions and tasks without you picking a model.
• Pick a specific model when you want the same model across several chats, or when your team recommends one for a certain kind of work.

Reasoning modes

Glean plans a task step by step and adapts as it works. On a Customer Key, choose a reasoning mode from the chat input:

• Adaptive: Picks the right mode for you. It starts with a fast response and automatically switches to deeper reasoning when your question needs it or requires tools that Fast mode doesn't support. Adaptive is the default mode when it is available.
• Fast: Provides quick answers for simple tasks. Tools from connected apps and Skills aren't available in this mode. Switch to Thinking if you need them.
• Thinking: Uses more reasoning for complex tasks and searches more thoroughly across your company's knowledge. It works well for multi-step questions that benefit from citations and detailed responses and supports the full set of tools, including tools from connected apps and Skills.
• Deep research: Appears in the mode picker only for Customer Key organizations where Deep Research is still a chat mode. It searches your company's knowledge and the web, then writes a citation-rich report. It's read-only and doesn't use tools that make changes in connected apps.

Glean Universal Model Key organizations don't choose Adaptive, Fast, or Thinking. Auto routes each question to a model and reasoning effort. If you pick a specific model, you can choose its reasoning effort from the model picker. Each response shows which models Glean used.

Everywhere else, Deep Research is a Glean Skill rather than a mode. Open the skill picker and select Deep Research, or type /deep-research and your question. An admin has to turn Deep Research on and give you Skills access. See Deep Research.

On a Customer Key, the mode controls how much Glean thinks, and the model controls which AI model responds. You set both from the chat input. To pick a model or let Glean choose one with Auto, see Which model should I use?. If you're an admin, see model choice.

If Assistant returns instructions instead of performing a connected-app action, and no Connect prompt appears, connected-app tools were not loaded for that turn. On a Customer Key, switch to Thinking. See Troubleshooting tools authentication.

Skills behave the same way on a Customer Key. In Fast mode, Glean doesn't load Skills at all, so a Skill won't run even if you reference it by name. Assistant answers without it and may report that the Skill isn't available to it. This is not a permissions issue. Switch to Thinking mode. On the Glean Universal Model Key, there's no Fast mode, so Skills can run with Auto or any model you pick. See Skills.

Where to use Glean

You can access Glean from multiple surfaces:

• Web app: Go to your organization's Glean URL or click New chat in the left navigation.
• Desktop app: A dedicated app for macOS and Windows with quick-chat access from anywhere on your machine. See Glean for Desktop.
• Mobile: Use Glean on the go. See Mobile.
• Browser extension: Access Glean from the sidebar on any webpage. See Extension sidebar.
• Glean companion: A floating overlay that provides context-aware assistance as you browse. See Glean companion.