DocumentationKnowledge and memory

Knowledge sources

Index selected content from GitHub, Notion, Slack, Linear, or a website, then grant the finished source to specific teams.

Agents can answer questions from the material your company already keeps in GitHub, Notion, Slack, Linear, or on a website. You decide what Hivy indexes and which teams may search it.

Knowledge is context, not a live tool

A connection lets an agent read or change data in an external service. A knowledge source indexes a chosen set of content for search. You may connect the same service both ways, but search and live actions remain separate.

Compare knowledge with connections

Video placeholder

Add your first knowledge source

Capture an admin opening Settings > Knowledge and choosing Add source. Show the provider, a narrow resource selection, the Teams field, and the source status after saving. End on the source list as indexing starts; every label and selected resource must remain readable.

Bring in the sources agents need

GitHub

Make selected repositories searchable when an agent needs code or engineering context.

Notion

Index chosen pages and databases without giving agents the whole Notion workspace.

Slack

Keep decisions from selected resources available after a Slack thread has gone quiet.

Linear

Let agents search work from the Linear teams you select.

Website

Discover a site, then choose the pages or sections agents may search.

Add a source in four steps

Workspace owners and admins add sources under Settings. If the content lives in an external service, connect that service before you start.

  1. 1

    Pick the source

    Choose a connected service or enter a website. A connected service reads through an existing workspace connection.

  2. 2

    Set a narrow scope

    Select only the repositories, pages, databases, Slack channels, Linear teams, or website paths that answer the questions you have in mind.

  3. 3

    Choose the teams

    Select each team whose agents may search this source. Hivy requires at least one team.

  4. 4

    Wait for indexing

    Hivy reads the chosen content in the background. Documents become searchable as the index fills.

Add a source

Keep each source narrow

Give the source a name that says what it contains, such as Engineering repositories or Support handbook. Include only the resources that belong under that name; tighter scope makes search results easier to judge and access easier to review.

For a website, enter the URL and run discovery before selecting individual pages or complete sections. With a connected service, Hivy lists the repositories, pages, Slack channels, or Linear teams that the connection can read.

Image placeholder

Select a focused source scope

Capture the Add knowledge source form after choosing a provider. The source name, resource picker, Teams field, and a short resource selection must fit in one readable frame. Use demo content.

Check the index before relying on it

The source list reports the source state and indexed document count. A large first sync may still run after its earliest documents become searchable, so check the count before treating the source as complete.

Choose View documents from the source menu to check indexed titles and links. The list should match the scope you meant to add.

Open Knowledge settings

Ask a question, not a search query

Tell the agent to search company knowledge, then ask the question in plain language. “What did we decide about the launch date?” works; so does “Find the support escalation policy.”

Hivy refreshes connected content on a schedule. Use the provider's connection when you need its current record or want the agent to change something there.

Read about agent sessions

Start with a question your team asks often

Add only enough material to answer that question, then grant the source to the team that owns the work. Test it in a session. If the answer lacks context, add the specific resource that holds the missing information.