Connect Your Knowledge Base to an AI Support Agent
Set up a live connection between your External KB and an AI agent platform so the agent can retrieve and answer questions from your published articles.
Connecting your External KB to an AI agent platform takes about ten minutes and requires no code. You'll generate a feed URL and API key inside your KB settings, paste those credentials into your AI agent platform, choose which products the agent can access, and run a test query to confirm the integration is live. Once the connection is active, your agent automatically uses your published articles as its knowledge source for every customer conversation.
Prerequisites
Before you begin, confirm the following:
- Published External KB content — You have at least one product with one or more published articles in your External KB. The agent indexes only published content, so a KB with only drafts won't return any results.
- AI agent platform account — You have an active account on the AI agent or chatbot platform you intend to connect. The platform must support an external knowledge source configured via a URL and API credentials (most modern platforms do).
- KB Admin permissions — You need Admin access to your KB workspace to generate API keys and manage integrations.
Connection Setup
In your KB dashboard, navigate to External KB → AI Agents, then click New Connection. This opens the connection wizard, where you'll configure each aspect of the integration.
Choose your AI agent platform from the list of supported integrations. If your platform isn't listed by name, select Generic REST API — this option works with any platform that accepts a feed URL and bearer token. The selection determines the exact format of the credentials page shown in the next step.
Click Generate API Key. The wizard displays two values:
- Feed URL — The endpoint your agent platform will call to fetch KB content (e.g.,
https://kb.yourapp.com/api/v1/kb/feed). - API Key — A bearer token that authenticates requests to your feed.
Copy both values and store them somewhere secure. The API key is shown only once — if you close this screen without saving it, you'll need to rotate the key and generate a new one.
Switch to your AI agent platform and open its knowledge source or data integration settings. Paste in the Feed URL and the API Key exactly as copied. Most platforms label these fields "Knowledge Base URL" and "API Token" or similar. Refer to your platform's documentation if you aren't sure where these fields are.
Back in your KB connection wizard, choose which products and topic categories the agent should have access to. You can grant access to your entire External KB or restrict the agent to specific products. Scoping the connection keeps the agent focused and prevents it from surfacing articles that aren't relevant to the support context where it's deployed.
Use the built-in Test Query field to send a sample question to the agent through the feed. The test panel shows the raw API response including matched articles, relevance scores, and any errors. Confirm that the response contains articles you'd expect for the question you asked. If the response is empty, verify that you have published articles matching the topic and that the correct products are scoped to this connection.
Once your test query returns the expected results, click Enable Connection. The agent platform will begin polling your KB feed on its configured schedule (see Training & Tuning for sync frequency details). Your agent is now live and will use your published KB articles to answer customer questions.
Feed URL and API Key Reference
Your Feed URL is a stable HTTPS endpoint that returns your published KB articles as structured JSON. It accepts query parameters so the agent platform can filter by product, category, or date range. The URL does not change unless you explicitly reset the integration — you only need to update your agent platform if you rotate your API key.
Your API Key is a bearer token that must be included in every request to the feed. It is scoped to the specific connection you created, meaning it only grants access to the products you selected during setup.
Finding Your Credentials Again
If you need to retrieve your Feed URL after the initial setup, go to External KB → AI Agents, locate your connection, and click View Details. The Feed URL is always visible here. The API Key is not re-displayed after generation — if you've lost it, rotate the key as described below.
Rotating a Compromised API Key
If your API key is exposed or you suspect unauthorized access, rotate it immediately:
- Go to External KB → AI Agents and open the affected connection.
- Click Rotate API Key and confirm the action.
- Copy the new API key.
- Update your AI agent platform with the new key.
The old key is invalidated instantly upon rotation — your agent will stop returning results until you update the platform with the new credentials.
Never expose your API key in client-side code, public repositories, or anywhere accessible outside your server environment. Requests to the KB feed should originate from your AI agent platform's backend, not from a browser or mobile app. A leaked key grants read access to all published articles in the scoped products, and could allow unauthorized parties to query your content.
Example API Request and Response
This is the format of a standard request to your KB feed. Your AI agent platform sends this request automatically — you don't need to build it yourself — but it's useful to understand the structure when debugging.
GET /api/v1/kb/feed?product=slack&format=json
Authorization: Bearer YOUR_API_KEYA successful response returns an array of article objects:
{
"product": "slack",
"total_articles": 3,
"articles": [
{
"id": "art_0f3a91c2",
"title": "How to Set Up Slack Notifications",
"body_snippet": "To configure notifications in Slack, open your workspace settings and navigate to Notifications. From there, you can customize alerts by channel, keyword, or activity type...",
"url": "https://help.yourapp.com/external/slack/notifications",
"topic_category": "Configuration",
"last_updated": "2025-06-10T14:22:00Z"
},
{
"id": "art_1b7e24d8",
"title": "Managing Slack Channel Permissions",
"body_snippet": "Channel permissions control who can post, invite members, and adjust settings within a given channel. Admins can set these at the workspace level or override them per channel...",
"url": "https://help.yourapp.com/external/slack/channel-permissions",
"topic_category": "Administration",
"last_updated": "2025-05-28T09:05:00Z"
},
{
"id": "art_2c9d57f4",
"title": "Troubleshooting Slack Integration Errors",
"body_snippet": "If your Slack integration shows a connection error, start by verifying that your OAuth token is still valid and that the app has not been removed from your workspace...",
"url": "https://help.yourapp.com/external/slack/integration-errors",
"topic_category": "Troubleshooting",
"last_updated": "2025-06-01T11:47:00Z"
}
]
}Each article object includes:
| Field | Description |
|---|---|
id | Unique identifier for the article |
title | The article's published title |
body_snippet | A truncated preview of the article body (first ~400 characters) |
url | The public URL of the published article |
topic_category | The category the article is filed under |
last_updated | ISO 8601 timestamp of the last publish event |
When setting up a connection, scope it to the specific products relevant to the support context where the agent is deployed. If you're embedding an agent on your Slack integration help page, scope the connection to the Slack product only. A focused agent gives more precise answers than one with access to your entire KB — and it's easier to audit and improve over time.