Model Context Protocol

MCP Tools for Elearning

Ask your AI assistant to work on Content Studio storyboards and media or configure AI Interactions for Storyline and HTML. Choose a product, then follow its connection guide.

Connect your AI assistant

MCP connects your AI assistant to Cluelabs. You choose which resources it can access and whether it can make changes.

Navigation

Connect your AI assistant

Choose your app, then follow its sign-in or API token instructions. Menu names may vary with your app version and your organization's settings.

Choose your Cluelabs product

AI Interaction

Connect AI Interaction with an API token

Use an account API token to inspect or edit Storyline and HTML interactions, including their instructions, knowledge base, variables, and optional web files. Choose the interactions and permission level your assistant needs.

Server address:

https://cluelabs.com/api/cluelabs/interaction/mcp

This connection uses Streamable HTTP with Authorization: Bearer YOUR_API_TOKEN. Create the token in Account Security; the editor displays your interaction ID without exposing a reusable editing credential.

After connecting, ask the assistant to call list_interactions. A successful list, even an empty one, confirms access. Seeing the catalogue alone does not.

Back to setup guides

AI Interaction

Create an AI Interaction API token

  1. Open Account Security and create an API token with a descriptive label and expiration.
  2. Select AI Interaction. Choose one interaction or all interactions you can access. Both Storyline and HTML resources appear in the picker.
  3. Choose read access (10), write access (20), or all permissions (250). A single-resource grant cannot exceed your current access. An all-resource grant is a ceiling; current ownership and sharing still limit every operation.
  4. Copy the secret when it is shown once, then add it as an Authorization header in your client. Revoke unused tokens in Account Security.

Creating interactions requires all-resource write or all-permissions scope. A single-interaction token cannot create another interaction. Whole-interaction deletion also requires current ownership; a collaborator with level 250 cannot delete the interaction.

Editors can disable logging or create a new table owned by the interaction owner. Below effective level 250, a changed positive table selection retains the existing binding while the rest of the save succeeds. At level 250 a replacement table must belong to the interaction owner.

Existing connections must be reconfigured manually with an account token and explicit interaction IDs.

Back to setup guides

AI Interaction

AI Interaction and ChatGPT

ChatGPT desktop: Use an API token

  1. Create your AI Interaction API token.
  2. In ChatGPT desktop, open Settings > MCP servers > Add server. Name it cluelabs-interaction, choose Streamable HTTP, and enter https://cluelabs.com/api/cluelabs/interaction/mcp.
  3. Save the server. To supply the token, edit your personal ~/.codex/config.toml (%USERPROFILE%\.codex\config.toml on Windows). Update this server's existing entry to match the example below; replace YOUR_API_TOKEN. Keep other entries.
[mcp_servers.cluelabs-interaction]
url = "https://cluelabs.com/api/cluelabs/interaction/mcp"
http_headers = { Authorization = "Bearer YOUR_API_TOKEN" }

Keep this file private: it contains your token. Select Restart in the MCP settings. In a local desktop chat, use /mcp to check the connection, then ask it to list your AI Interactions. A successful result confirms token access.

This direct desktop connection uses the configuration shared with Codex. ChatGPT web plugins have separate authentication settings. See OpenAI's desktop MCP setup and header configuration.

Ask ChatGPT to call list_interactions. AI Interaction uses API tokens; the OAuth setup for authenticated ChatGPT web plugins does not configure this token connection.

Back to setup guides

AI Interaction

Connect AI Interaction to Claude Desktop

Connector with Request headers

  1. Create an AI Interaction API token.
  2. In Claude Desktop, open Customize > Connectors, select + Add > Add custom connector, and enter Cluelabs AI Interaction and https://cluelabs.com/api/cluelabs/interaction/mcp. Select Continue and review the detected settings.
  3. Under Authentication, choose No sign in. Open Request headers, select authorization, enter Bearer YOUR_API_TOKEN with your token, and mark it Required. Include the space after Bearer; Claude sends the value exactly as entered.
  4. Select Add. In a new chat, open + > Connectors and enable AI Interaction.

On a team account, an owner adds the connector in Organization settings > Connectors > Add > Custom; choose Web if asked. Dialog layouts vary by organization.

Header authentication is available to a limited set of organizations. See Claude's current header instructions.

Desktop alternative when Request headers is unavailable

With Node.js installed, open Settings > Developer > Edit Config and merge this entry into claude_desktop_config.json. Replace the token placeholder, save, completely quit and reopen Claude Desktop.

{
    "mcpServers": {
        "cluelabs-interaction": {
            "command": "npx",
            "args": [
                "-y",
                "mcp-remote",
                "https://cluelabs.com/api/cluelabs/interaction/mcp",
                "--header",
                "Authorization:${AUTHORIZATION}"
            ],
            "env": {
                "AUTHORIZATION": "Bearer YOUR_API_TOKEN"
            }
        }
    }
}

The helper's own documentation explains header forwarding and the space-free argument used on Windows.

Claude on the web

The local Desktop helper is unavailable on the web. If your web connector has no Request headers setting, this Interaction connection cannot authenticate there; use a supported token-capable client. On shared connectors, the token's scope applies to everyone using that credential.

Ask Claude to call list_interactions and return the IDs and modes.

Back to setup guides

AI Interaction

Connect AI Interaction to Gemini CLI

Merge this entry into the mcpServers map in your Gemini CLI ~/.gemini/settings.json. Replace the token placeholder, save, and start a new CLI session. Use /mcp to check connection status, then ask it to call list_interactions.

{
    "mcpServers": {
        "cluelabs-interaction": {
            "httpUrl": "https://cluelabs.com/api/cluelabs/interaction/mcp",
            "headers": {
                "Authorization": "Bearer YOUR_API_TOKEN"
            }
        }
    }
}

See Gemini CLI's HTTP and custom-header configuration.

Gemini Enterprise

This Interaction release adds no Gemini Enterprise OAuth setup. Use the Gemini CLI token configuration above where available. Content Studio's administrator OAuth instructions do not configure AI Interaction.

Back to setup guides

AI Interaction

Connect AI Interaction to Antigravity

In Antigravity IDE, open the MCP servers panel and its raw configuration. Merge this remote server into mcp_config.json, replace the token placeholder, save, and refresh the server connection. Ask it to call list_interactions.

{
    "mcpServers": {
        "cluelabs-interaction": {
            "serverUrl": "https://cluelabs.com/api/cluelabs/interaction/mcp",
            "headers": {
                "Authorization": "Bearer YOUR_API_TOKEN"
            }
        }
    }
}

The Antigravity documentation specifies Streamable HTTP and custom headers for remote servers.

Back to setup guides

AI Interaction

Connect AI Interaction to Visual Studio Code

Open your VS Code .vscode/mcp.json or user MCP configuration and merge the server below. Replace the token placeholder in your private configuration, save, start the server, enable its tools in the chat, and ask it to call list_interactions.

{
    "servers": {
        "cluelabs-interaction": {
            "type": "http",
            "url": "https://cluelabs.com/api/cluelabs/interaction/mcp",
            "headers": {
                "Authorization": "Bearer YOUR_API_TOKEN"
            }
        }
    }
}

Keep populated credentials out of source control. This is the VS Code configuration format; other harnesses may use different configuration and credential handling. See the VS Code MCP reference.

Back to setup guides

AI Interaction

Connect AI Interaction to Kiro

In Kiro IDE, use the command palette to open the user MCP configuration at ~/.kiro/settings/mcp.json. Merge this server, replace the token placeholder, and save to reconnect. Ask Kiro to call list_interactions.

{
    "mcpServers": {
        "cluelabs-interaction": {
            "url": "https://cluelabs.com/api/cluelabs/interaction/mcp",
            "headers": {
                "Authorization": "Bearer YOUR_API_TOKEN"
            }
        }
    }
}

See Kiro's remote-header configuration. Workspace configuration can override user entries.

Back to setup guides

AI Interaction capabilities

Interaction tool reference

Choose IDs from list_interactions. Existing-resource calls require a positive interaction_id. Read operations require 10, ordinary writes require 20, and deletion requires 250 plus actual ownership. The storage-zone helper accepts any valid supported account token.

ToolInputsResultsAccess
list_interactions{}interactions10
create_interactionname, folder_id, zone_gcp, settings, contentinteraction20
get_interactioninteraction_idinteraction10
update_interactioninteraction_id, settings, contentinteraction20
rename_interactioninteraction_id, nameinteraction_id, name, saved_at20
delete_interactioninteraction_idinteraction_id, deleted250
get_storage_zones{}zones0
list_interaction_filesinteraction_idinteraction_id, display_url, mode, web_mode_active, files, images10
get_interaction_file_contentsinteraction_id, file_pathfile_path, content, size10
update_interaction_fileinteraction_id, file_path, contentinteraction_id, saved, saved_at20
upload_interaction_imageinteraction_id, filename, content_base64interaction_id, name, path, mime, width, height, resized20
delete_interaction_imageinteraction_id, image_pathinteraction_id, deleted20
get_interaction_display_urlinteraction_idinteraction_id, display_url, mode, web_mode_active10
get_interaction_contextinteraction_idinteraction_id, mode, settings, content, display, editing_surface10

Creation and configuration saves complete synchronously, including required knowledge-base processing. File writes validate code and provision missing companions without overwriting edits; images retain existing sanitization and size limits. A display URL does not establish successful HTML playback.

Example workflows

  • Read scope: “List my interactions with IDs and modes.”
  • Wildcard write scope: “Create a Storyline interaction named Safety practice,” or “Create an HTML interaction named Safety practice web.”
  • Write scope for the identified resource: “For interaction 123, update these instructions, knowledge base, and variables.” Omitted settings stay unchanged; supplied variables replace the list.
  • Write scope: “Read interaction 123's style.css, then apply these changes to that file.”
  • All permissions and ownership, only after an explicit request: “Permanently delete my interaction 123 and its workspace.” Independent linked log tables remain.

Frequently asked questions

Does this work in both modes?

Yes. Lifecycle and configuration tools support Storyline and HTML. Workspace edits leave the active mode unchanged.

Which interactions can the token reach?

Single-resource tokens reach that ID within your current access. All-resource tokens cover currently accessible resources, including later creations and shares, capped by the token level.

Why is the workspace empty?

Reads list only existing files. A Storyline interaction can have no web workspace. Reading a missing file returns file_not_found; deleting a missing image returns image_not_found. Storage outages are reported separately.

How do I authenticate and select a resource?

Use an account token in the Authorization header and a returned interaction ID in tool arguments. Browser sessions and public runtime URLs do not authenticate MCP editing.

What happens after an incomplete save or deletion?

Inspect the identified resource with get_interaction or list_interactions. Do not automatically repeat creation. Deliberately update the known ID, or retry deletion only if the owned interaction remains.

Get Started

Start Building with MCP Today

Choose your product and app, follow its supported authentication instructions, and ask the assistant to list the resources your connection can access.

Choose your setup guide