Honen has separate connections for importing source material, letting its assistant call external tools, and letting another AI client work with Honen. Choose the connection for the direction you need.
| Connection | Purpose | Direction |
|---|---|---|
| Content source | Import selected documents and refresh their Knowledge Base copies | Source system into Honen |
| External MCP server | Give the Honen assistant tools from another system | Honen into external tools |
| Honen MCP or Connector API | Let an approved client use Honen commands | External client into Honen |
| Agent channel | Deliver a configured Honen Agent through chat or an embed | Honen and the channel audience |
Start with Import files and web links, GitHub sync, or Connect an AI client.
1. Bring content into the Knowledge Base
Connect Google Drive, OneDrive, SharePoint, Notion, Confluence, Document360, GitHub Documentation, or Granola, or import website pages and YouTube transcripts. Availability depends on workspace configuration and provider setup. Each adapter supports its own content types and scope; a selected folder imports supported descendants subject to crawl and processing limits.
Use Content sources to connect an account or enter the credential the provider requires. Return to the Knowledge Base add menu to select content. Websites and YouTube use URLs; Granola uses an API key. Follow the import status to see when the selected content is ready.
Manual Knowledge Base imports accept supported PDF, Word and PowerPoint documents, images, and video up to 5 GB per file. Markdown and plain-text imports accept up to 5 MB of text. This is the Knowledge Base import limit; other upload surfaces have their own limits. Supported screen recordings can be processed into visual walkthroughs. Review the generated steps and screenshots before using them.
Refresh timing and deletions
Connected sources refresh on their supported schedules. Timing depends on the provider and the amount of content being processed; website sources have a six-hour minimum scheduled refresh interval. Choose Refresh when you want to check for a recent change, then open the updated document.
Refresh uses the credential stored with the source connection. Personal connections use the linked owner's account; workspace-scoped connections use the shared workspace credential. The provider must still allow that credential to read the source. Honen's own sharing controls determine who can read the imported copy.
A complete folder refresh also reconciles documents removed from the source. Directly linked documents retain their last imported copy when source access changes. Manage the Honen copy and its sharing settings when retiring material.
Source imports are one-way: editing the Knowledge Base copy does not push that edit to the connected source, and a later source update can replace local edits. This is separate from external MCP tools, which may be authorized to write to the original system.
Images and scanned material
Honen copies supported referenced images into workspace storage during sync. Use accessible source assets and open the imported document to review the images alongside their text.
Supported scanned documents go through text extraction. Check the extracted text, especially tables and image-heavy pages, before using it as course material.
Course materials are snapshots
Attaching a Knowledge Base document to a course captures a version snapshot. The materials panel can flag a newer source version and offer Refresh from KB. That refresh replaces the attached material; it does not automatically rewrite or publish existing lessons and questions. Review source changes and revise the course separately. See Organize and use source material.
2. Let Honen call external tools
The MCP catalog includes Linear, Notion, Granola, HubSpot, Slack, Google Calendar, Google Chat, Google Drive, Gmail, and Google Contacts. These connections add tools to the assistant; choose a native content source when you want a synced Knowledge Base collection. Available operations depend on the provider, your authorization, and the selected connection.
You can also configure a custom MCP server using a supported remote transport and authentication flow. Complete its authentication flow, inspect the available tools, and try the operation you want to use.
Before using an external tool in an automation, confirm the connection is active and that the Agent can access the required operation. Review the run result to see the returned data and completed work.
3. Connect an outside AI client to Honen
- Personal sign-in: the client uses the signed-in person's Honen permissions, further limited by its granted OAuth scopes and the tool's access checks.
- Scoped Connector: an administrator configures permitted capabilities and resources, such as read access to a specific folder. An enabled Knowledge Base capability with no resource restrictions can cover the whole workspace.
The OAuth-protected Streamable HTTP endpoint is https://honen.com/api/mcp/mcp. Discovery is available at https://honen.com/.well-known/mcp, and the Server Card at https://honen.com/.well-known/mcp/server-card.json describes the supported surface. Clients supporting Honen's MCP App can display its interactive course preview; other clients may only expose the tools and returned links.
Follow Connect an AI client for the setup screens or API and OpenAPI for HTTP commands. Both require the relevant permissions; an API credential is not an LMS or xAPI credential.
4. Deliver an Agent through a channel
An Agent shares its instructions and Connector policy across its configured Slack, Microsoft Teams, iMessage, and website-embed deployments. Provider credentials, identity modes, and routing are configured per deployment. Follow the selected channel’s installation guide to activate the deployment.
Review access and execution evidence
Knowledge Base path rules can narrow access, including a read-only document inside a writable folder. With rules configured, unmatched items are denied; with the flat policy, empty resource allowlists mean no resource restriction. Review the saved policy rather than relying on the Agent's written instructions.
Imported content follows Knowledge Base sharing permissions. Connecting a source does not by itself publish it, but importing into a shared folder can make that copy available to the folder's audience. Check the destination before importing private content.
Use integration audit records, document version history, and Agent logs to follow changes and review recorded conversations and tool activity.
Disable or revoke credentials you no longer need and review provider-specific revocation behavior. Credential lifetimes differ: a Connector can be created without an expiry, while short-lived delegated tokens have their own limits. Set an expiry where offered and keep ownership of each credential clear.
Identity and LMS delivery
Use OIDC or Signed JWT to connect supported identity flows. For district identity or rostering requirements, confirm the supported setup with your administrator before planning a rollout. Follow Single sign-on and identity to choose and configure the learner sign-in flow.
For learning delivery, use LMS Integrations. LTI launches, Deep Linking, roster sync, and grade passback depend on the selected platform's capabilities and configuration. SCORM, Google Classroom, and xAPI/cmi5 have separate setup and verification workflows.
Keep sources connected
Review connection and sync status when the source account, ownership, or permissions change. Reconnect the authorized account when prompted, choose Refresh, and open an updated document to confirm the current source and images.
For workforce sign-in, start with Single sign-on and identity, then follow OIDC setup or Signed JWT setup. Use Governance to align templates, Rules, Skills, and publishing approvals with your team's responsibilities.