AI API Access

AI API connects an external AI assistant to one iTasks workspace. The assistant can access only the selected workspace data: projects, members, tasks, comments, time logs, and HelpDesk workflows.

1. Requirements Before Connecting

• Only a user with the OWNER role can configure access.
• AI API requires an active PRO subscription on this workspace.
• If PRO expires, discovery, MCP, OpenAPI, and REST stop working until the subscription is restored. Workspace data is not deleted automatically; plan behavior is covered in Subscription, Plans, and Limits.

2. How to Enable AI API

1. Open Administration → Workspace settings.
2. In the AI access section, select AI API.
3. Create an access token if it has not been created yet.
4. Enable AI access for the workspace.
5. Copy the connection link and send it to the AI assistant.

The link already contains the workspace token, so the AI client does not need to choose or send a workspaceId. All requests are automatically scoped to one workspace.

3. Extra Bearer Protection

The URL token inside the link grants access to the workspace AI API by itself. For real integrations, create a Bearer token and enable extra protection: the client must then send Authorization: Bearer <token> together with the connection link.

The full Bearer token is shown only once after creation or rotation. If the client loses it, the owner must rotate the token in this section. Link-only access should be left enabled only for AI systems that technically cannot send an authorization header.

4. Available Connection Options

The interface shows several URLs for different client types:

• Connection link — the recommended discovery address for most assistants.
• MCP URL — Model Context Protocol access.
• OpenAPI schema — an OpenAPI 3.1 description of the REST API for clients that can import a schema.
• REST base URL — the base JSON REST API URL for manual setup.

5. Working Scenarios: Codex and Claude Code

For a quick discovery check, send the assistant the connection link. For manual MCP client setup, use the MCP URL. If extra Bearer protection is enabled, the MCP URL must be used together with the Bearer token, which iTasks shows only once after creation or rotation.

# If extra Bearer protection is enabled, first set ITASKS_AI_BEARER_TOKEN in your shell.
codex mcp add itasks --url "<MCP_URL_FROM_ITASKS>" --bearer-token-env-var ITASKS_AI_BEARER_TOKEN
codex mcp list

codex mcp add itasks --url "<MCP_URL_FROM_ITASKS>"
codex mcp list

Connect to iTasks with this link:
<CONNECTION_LINK_FROM_ITASKS>

Do not change anything yet. Read the discovery/MCP capabilities, check projects, members, and task statuses. Briefly explain what you can do in this workspace and which actions need my explicit confirmation.

Use the iTasks MCP server `itasks`. In the "<PROJECT_NAME>" project, prepare and launch a landing page about "<TOPIC>".

First run safe read operations: projects, members, and task statuses. Then create a task set with clear titles, Taskbot rich text HTML descriptions, priorities, due dates, and assignees selected only from the discovered members. Do not delete or close existing tasks. At the end, return the created task list, assignees, due dates, and assumptions.

claude mcp add --transport http --header "Authorization: Bearer <ITASKS_BEARER_TOKEN>" itasks "<MCP_URL_FROM_ITASKS>"
claude mcp list

claude mcp add --transport http itasks "<MCP_URL_FROM_ITASKS>"
claude mcp list

Connect to iTasks with this link:
<CONNECTION_LINK_FROM_ITASKS>

Do not change anything yet. Read the discovery/MCP capabilities, check projects, members, and task statuses. Briefly explain what you can do in this workspace and which actions need my explicit confirmation.

Use the iTasks MCP server `itasks`. In the "<PROJECT_NAME>" project, prepare and launch a landing page about "<TOPIC>".

First run safe read operations: projects, members, and task statuses. Then create a task set with clear titles, Taskbot rich text HTML descriptions, priorities, due dates, and assignees selected only from the discovered members. Do not delete or close existing tasks. At the end, return the created task list, assignees, due dates, and assumptions.

6. What the Assistant Can Work With

The current implementation exposes workspace workflows to AI tools: reading projects, members, tasks, and statuses; task search; creating, updating, and deleting tasks; changing status, workflow, and assignee; working with subtasks, comments, and time logs; viewing HelpDesk channels, conversations, and messages; adding external replies, internal notes, and task links for conversations.

7. Technical Notes for Client Setup

• The main REST exchange format is JSON.
• Formatted text fields for tasks, comments, and HelpDesk messages are sent as HTML.
• AI API has workspace-scoped rate limits. When the limit is exceeded, the API returns 429, so clients should avoid tight polling.
• The connection link, MCP, OpenAPI, and REST belong to one workspace. Another workspace requires its own owner-configured access.

Related Guides

General workspace controls are covered in Workspace Settings. Owner and admin roles are covered in Members and Roles. If AI needs to work with HelpDesk, configure bots and support channels first.