Salesforce Knowledge

The Salesforce Knowledge integration ingests knowledge articles from your Salesforce org into Kapa. Articles are fetched through a Salesforce account connection, so both public and internal knowledge bases are supported.

note

An older connection method that scraped public Salesforce sites is deprecated and will stop working in the future. If you have an existing source using it, see Migrating from public site scraping.

Check what you expose

The account connection can read everything the connected Salesforce user can read, including internal-only knowledge. What Kapa ingests is controlled by the publishing channels you select, and answers in a public-facing project surface the ingested content to your end users. If your project is public-facing, only select channels whose articles are intended for a public audience. See Publishing channels.

Prerequisites

• A Salesforce org (Production) with permission to install an External Client App
• A Salesforce user with read access to the knowledge articles you want to ingest. Kapa ingests exactly what this user can read, so we recommend a dedicated integration user whose permissions match what you want ingested.

Data ingested

• Article titles and content (the fields on each article's page layout)
• Article URLs
• Topics
• Publishing channel visibility
• Publication dates

Only the latest published version of each article is ingested. Drafts and archived articles are never included.

Ownership

The Salesforce Knowledge source is tied to the Kapa user who completes the OAuth connection. Only that user can modify the source configuration. Other team members can view the connection details but cannot change them.

If the connecting user's Kapa account is later deleted, the source becomes locked and cannot be reconfigured. In that situation, create a new Salesforce Knowledge source.

Setup

Step 1: Install the Kapa.ai External Client App in your Salesforce organization

This is already done if you use the Salesforce Cases integration.

1. Open up this link to install the Kapa.ai External Client App
2. Install the application, choosing Install for All Users, or Install for Specific Profiles if only certain profiles should be able to create Salesforce sources
3. To ensure that the ingestion continues to work, make sure to mark the Refresh Token Policy -> Refresh token is valid until revoked in the App Authorization section of the app

Step 2: Connect your Salesforce org

1. Go to the Sources tab in the Kapa platform
2. Click Add new source and select Salesforce Knowledge
3. Click Connect Salesforce
4. A Salesforce login window opens. Sign in as the intended integration user. The consent screen silently reuses your browser's active Salesforce session, so log out first or use a private window if you are logged in as someone else (see Troubleshooting)
5. Approve the Kapa External Client App when prompted
6. The window closes automatically and the Connect step shows a Connected to Salesforce confirmation

Step 3: Configure and save

1. Publishing channels (required): only articles visible on at least one selected channel are ingested. Read Publishing channels before enabling non-public channels on a public-facing project
2. Public article base URL (optional): your public Experience Cloud site, used to build citation links (see Public article base URL)
3. Topics (optional): include or exclude articles by topic
4. Click Save to start the initial ingestion

Publishing channels

Salesforce publishes each article to one or more channels. Kapa filters ingestion by the channel flags on each article:

Channel	Salesforce meaning
Public Knowledge Base	Articles visible to anonymous visitors
Internal App	Internal-only articles for logged-in Salesforce users
Customer Site	Articles for logged-in customers on your Experience Cloud site
Partner Site	Articles for logged-in partners on your Experience Cloud site

Kapa relies on these flags to decide what is safe to ingest, but the flags are technically optional in Salesforce: an article can be published while carrying no channel flags at all (articles created through the Salesforce API often are), and such articles match no channel selection and are never ingested. Make sure the articles you want ingested actually carry the intended channel flags. During setup, Kapa warns you if no articles match your current selection, so an empty configuration is caught before saving.

Note that the Public Knowledge Base flag declares an article intended for anonymous visitors; it does not by itself make the article reachable on your public site. Public availability is controlled by separate Salesforce settings (site access, guest permissions, article sharing and visibility). These settings live deep in Salesforce and change over time, so refer to the Salesforce documentation or the team that manages Salesforce at your company when in doubt. See Public article base URL for how Kapa verifies what visitors actually get.

Avoid exposing internal knowledge

Every channel except Public Knowledge Base is protected behind a login in Salesforce: Internal App articles are usually written for an internal audience, and Customer Site and Partner Site articles are typically not meant for public viewing. Ingesting them makes their content available in your assistant's answers. Only enable these channels on a public-facing project if the articles are genuinely intended for a public audience. Kapa asks for explicit confirmation before applying such a selection.

Public article base URL

Kapa cites the articles an answer is grounded in. By default those citations link to the article's Salesforce URL, which requires a Salesforce login: fine for an internal deployment, but not something your end users can open on a public-facing one.

If your knowledge base is also published on a public Experience Cloud site, enter its base URL (usually ending in /s/, for example https://yourcompany.my.site.com/help/s/ on the default Salesforce domain, or https://support.yourcompany.com/s/ on a custom domain). Citations for publicly visible articles then link to your public site instead, so anyone reading an answer can follow them. For public-facing projects, set this URL whenever a public site exists.

Kapa validates the URL when you enter it:

• A green check means the site is reachable and verified to serve articles from your connected Salesforce org
• A warning icon means the site is reachable but could not be verified, because no article is published to the Public Knowledge Base channel yet. You can still save; verification activates once a public article exists
• An error means the URL is not an accessible Salesforce site, or it did not return articles from your connected Salesforce org when checked

If the check itself cannot run (for example, a temporary network issue), the field shows "We couldn't validate this URL" and saving is blocked. Edit the URL to retry the check, or clear the field to save without a citation URL.

Migrating from public site scraping

Sources using the deprecated public scraping method keep working for now, but new sources always use the Salesforce account connection. To migrate:

1. Create a new Salesforce Knowledge source and connect your Salesforce account (no new Salesforce setup is needed if the Kapa.ai app is already installed)
2. Select the Public Knowledge Base channel to ingest the same articles the public scraper covered
3. Set the Public article base URL to your public site so citations keep linking there
4. Once the new source has completed its first sync, delete the old source

Configuration options

Option	Description	Default	Required
Publishing channels	Only ingest articles visible on selected channels	Public Knowledge Base	Yes
Public article base URL	Public site used to build citation links	None	No
Topics	Include or exclude articles by topic	All articles	No

Troubleshooting

Connecting

• OAuth window closes with an error: ensure your Salesforce user has permission to authorize the Kapa External Client App and that pop-ups are not blocked in your browser.
• "Admin approval required" during consent: your Salesforce admin needs to install the Kapa.ai app or approve it for your user.
• Connected as the wrong user: the consent screen silently reuses your browser's active Salesforce session. To connect a specific integration user, log out of Salesforce first or run the connect flow in a private browser window.
• Connection stops working after connecting other sources: Salesforce limits the number of active tokens per user per connected app; connecting many sources or reconnecting repeatedly with the same Salesforce user silently revokes the oldest token. Use a dedicated integration user per source type.

Articles and channels

• Draft or archived articles are missing: this is by design. Only the latest published version of each article is ingested; drafts and archived articles are excluded regardless of channel selection.
• "No articles match the selected publishing channels": the connection is fine. No article carries the channel flags you selected. Check the article's channel visibility in Salesforce (the Visible In... checkboxes). Articles created through the Salesforce API often carry no channel flags at all.
• Channel visibility not editable in Salesforce: the channel flags can only be changed on a draft version of an article: edit as draft, set the visibility, and publish. If the visibility fields are not shown in your Salesforce UI, work with your Salesforce admin to surface them.

Public article base URL

• Validation fails or warns on a site that works in your browser: "working" while logged into Salesforce is not the same as publicly accessible: an active Salesforce session makes non-public sites look public. Open the site in a private browser window to see what visitors (and Kapa's check) actually get, then match the specific message below.
• "This URL isn't an accessible Salesforce Knowledge base": check that you entered the full community base path including the trailing /s/ (custom path prefixes like https://community.yourcompany.com/acme/s/ are fine), and that the site is live and publicly accessible without login.
• "This URL doesn't seem to return your connected Salesforce org's articles": either the URL points at a different Salesforce org's site, or visitors cannot view your Knowledge articles on it. Verify by opening an article link in a private browser window. If it does not render, work with your Salesforce admin on guest access to Knowledge articles, since several Salesforce settings gate it (site access, guest permissions, article sharing and visibility).
• "Could not confirm it belongs to your Salesforce org": publish at least one article to the Public Knowledge Base channel; verification activates automatically.

Topics

• Empty topics dropdown: your org has no Topics (they are part of Experience Cloud). This is normal: channel filtering works independently of topics.