Connections
The Connections tab allows you to connect external data sources to your knowledge base. Connected sources are automatically indexed so your agents can search and retrieve content without copying data into the platform.
Connectors are organized into four categories: Cloud Storage & Documentation, Databases, Business Applications, and Cloud Platforms.
Cloud Storage & Documentation
Link Drive so agents can search Docs, Sheets, and files you allow. Content is indexed into the knowledge base; originals stay in Google.
Connect SharePoint Online sites and document libraries. Team files become searchable without copying them out of M365.
Sync spaces and pages from Confluence Cloud or Data Center. Wikis, specs, and runbooks stay available to your project AI.
Ingest structured product documentation from Zoomin. Ideal for manuals, release notes, and technical publications.
Databases
Attach MySQL or MariaDB with host, database, and credentials. Choose tables or views to index for grounded answers.
Connect SQL Server on-premises or on Azure SQL. Index only the schemas and objects you authorize for search.
Connect Oracle Database with your service account. Bring selected tables into the knowledge base for AI retrieval.
Use a connection string for Postgres, Redshift, and more. Use when there is no dedicated connector tile for your warehouse.
Connect Snowflake through the Generic SQL tile with a username and password, or with Microsoft Entra ID (OAuth client credentials).
Connect a Databricks SQL warehouse with a personal access token.
Connect Hadoop (Hive) with an LDAP, Kerberos, None, or No SASL login. SSL is available over the HTTP transport.
Business Applications
Cloud Platforms
Authentication Matrix
Every connector signs in with OAuth, Basic authentication (username and password), or a token, key, or key file.
| Connector | Authentication | What you provide |
|---|---|---|
| Google Drive | OAuth | Sign in with your Google account. No other fields. |
| Microsoft SharePoint | OAuth | Sign in with your Microsoft account and approve access. Your Microsoft admin may need to approve the permissions first. |
| Confluence Cloud | OAuth | Accept the permissions on Atlassian's screen. No other fields. |
| Confluence On-Premise / Data Center | Bearer token | Confluence Server URL and a bearer token. |
| Zoomin | Bearer token (optional) | API URL and Web URL. Without a token, only public pages can be read. |
| MySQL | Basic | Host, port, database, username, password, and tables. |
| Microsoft SQL | Basic | Host, port, database, schema, username, password, and tables. |
| Oracle | Basic | Host, port, database or service name, username, password, schema, and tables. |
| Snowflake | Basic or OAuth (client credentials) | A connection string, with either a username and password or the OAuth parameters. |
| Generic SQL | Basic, or OAuth for Snowflake | A connection string with either the username and password (for example Postgres or Redshift) or the Snowflake OAuth parameters, and the tables to index. |
| Databricks | Personal access token | Workspace host, token, HTTP path, catalog, and schema. OAuth is not supported. |
| Hadoop (Hive) | LDAP (Basic), Kerberos, None, or No SASL | Host, port, and database. A username and password are needed only for LDAP. |
| Odoo | Username and API key | Server URL, database name, username, and API key. |
| Google Cloud Platform | Service-account key file | A service-account JSON key. Access is controlled by the IAM roles on that service account. |
How each method works
- OAuth. You sign in on the provider's own page (Google, Microsoft, or Atlassian) and approve access. EKB never sees your password and keeps only an access token. Snowflake can also use OAuth (client credentials), which you set up in the connection string.
- Basic (username and password). You type the username and password into the connection form. For Snowflake and Generic SQL, they go inside the connection string. Hadoop (Hive) uses Basic when the login method is LDAP.
- Token or key. You create a bearer token, personal access token, or API key in the other system and paste it in.
- Key file. Google Cloud Platform uses a service-account JSON key that you paste into the form.
Best Practices
Connector Setup
- Use Specific Folders — Connect specific folders rather than entire accounts.
- Set Appropriate Schedules — Balance content freshness with system load.
- Monitor Sync Status — Regularly check sync success and failures.
- Test Connections — Verify connections work before relying on them in production.
Security
- Use Service Accounts — Prefer service accounts over personal accounts for stability and security.
- Limit Permissions — Grant only the permissions necessary for indexing.
- Regular Audits — Review connected accounts periodically and remove unused connections.
- Secure Credentials — Never share connection credentials.
Performance
- Selective Sync — Sync only the content your agents actually need.
- Schedule Wisely — Avoid scheduling syncs during peak usage times.
- Monitor Resources — Watch for performance impacts after enabling new connectors.
- Optimize Queries — For database connectors, use efficient queries to limit the data retrieved.
Troubleshooting
Connection failures
Problem: Connector fails to connect.
Solutions:
- Verify credentials are correct.
- Check network connectivity.
- Ensure the required permissions have been granted.
- Review connector-specific requirements.
Sync failures
Problem: Content is not syncing.
Solutions:
- Check sync schedule configuration in the Sync & Schedule tab.
- Verify the connector is still connected in the Connections tab.
- Review sync logs for errors.
- Try a manual sync using the Sync Now button to test the connection.
Missing content
Problem: Expected content is not appearing in the Knowledge Base.
Solutions:
- Verify the sync completed successfully.
- Check folder or query configuration for the connector.
- Ensure the content matches the sync criteria and filters.
- Review sync filters and settings.