Skip to content

Using NodeKAT

NodeKAT answers questions in plain language, drawing from the tools your organization has connected — Slack conversations, GitHub code, Jira tickets, Notion pages, Google Drive files, and ingested documents.

Ask naturally, just like you would ask a colleague. There’s no special syntax or command language to learn.

Examples of good questions:

  • “What did we decide about the migration timeline in Slack last week?”
  • “Where is the get_api_key function defined, and who calls it?”
  • “What’s the status of the login page redesign ticket?”
  • “Summarize the discussion around the on-call rotation in the #incidents channel”
  • “What does our onboarding doc say about setting up a dev environment?”
  • Be specific — mention names, dates, channels, or ticket numbers when you know them (“Jira ticket NKT-42” rather than “that ticket”)
  • Ask one thing at a time — a focused question returns more precise results
  • Mention the source — if you know where the answer lives (Slack, GitHub, a ticket), say so
AreaWhat It Searches
SlackMessages, threads, and channel activity (past and current)
GitHubCode, symbol definitions and references, repositories
JiraTickets, issues, overdue items, and recent activity
NotionPages, databases, and internal documentation
Google DriveDocs, Sheets, Slides, and PDFs
Knowledge baseIngested documents (entity extraction, embeddings, knowledge graph)

Every question is processed through a multi-phase workflow:

  1. Classify — NodeKAT determines what type of question you’re asking
  2. Route — It selects the tools that are relevant to your question
  3. Execute — It searches your connected tools in parallel
  4. Review — It checks whether the results fully answer the question
  5. Synthesize — It writes a final answer, citing the sources it used

Answers include a Sources section listing the documents, messages, tickets, or files the answer is based on. Use the sources to verify details and open the original material.

NodeKAT can only answer from content it has access to. If a message was sent in a channel the integration can’t read, or a document wasn’t shared, it won’t appear in answers — even if it’s in your tool.

  • No actions — NodeKAT is read-only. It cannot send messages, create tickets, push code, or modify files
  • No real-time awareness — it knows what’s in your connected tools and knowledge graph; it may not reflect changes from the last few minutes until the next sync
  • No guessing — if the information isn’t in a connected source, NodeKAT will say so rather than invent an answer

Find a decision:

“What did we agree on regarding the API rate limit in the #backend channel?”

Trace code:

“Find the definition of generate_token and all the places it’s referenced”

Track work:

“What Jira issues are overdue this week?”

Search documentation:

“According to our onboarding page, what are the steps to request access to staging?”

  • FAQ — common questions about using NodeKAT
  • Privacy & Data — how NodeKAT accesses and handles your data