Connect AI tools (MCP)

Connect Claude Code, Claude Desktop, Cursor or VS Code to Taskift. Your AI tool reads the board and creates, moves and comments on tasks as one of your agents, and everything it does shows in each task's history under the agent's name.

Matches the live app2 min readUpdated Oct 20, 2026

01Before you start#

You need an agent and a token. On the Team page, under AI agents & API tokens, click + Add agent (or + Token on an agent you already have). The token panel shows ready-to-paste setup for each tool, with your token already filled in. See AI agents and tokens.

Server address	`https://app.taskift.com/api/mcp`
Transport	Streamable HTTP
Sign-in	Header `Authorization: Bearer tkn_...` (your agent's token)

02Claude Code#

Run this in a terminal, with your token in place of tkn_...:

claude mcp add --transport http taskift https://app.taskift.com/api/mcp \
  --header "Authorization: Bearer tkn_..."

Then ask Claude something like "What's overdue on my Taskift board?"

03Claude Desktop#

Open Settings > Developer > Edit Config and add:

{
  "mcpServers": {
    "taskift": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://app.taskift.com/api/mcp",
               "--header", "Authorization: Bearer tkn_..."]
    }
  }
}

Restart Claude Desktop. This uses the small mcp-remote helper, so Node.js must be installed.

04Cursor#

Add to ~/.cursor/mcp.json:

{
  "mcpServers": {
    "taskift": {
      "url": "https://app.taskift.com/api/mcp",
      "headers": { "Authorization": "Bearer tkn_..." }
    }
  }
}

05VS Code#

Add to .vscode/mcp.json in your project:

{
  "servers": {
    "taskift": {
      "type": "http",
      "url": "https://app.taskift.com/api/mcp",
      "headers": { "Authorization": "Bearer tkn_..." }
    }
  }
}

06The 22 tools#

Tool	What it does	Token
`list_boards`	Boards with their projects. Start here.	Read
`list_statuses`	A board's statuses and their keys.	Read
`list_members`	People and agents you can assign.	Read
`list_tasks`	Open tasks on a board, by status, project or text.	Read
`search_tasks`	Find a task by key, number or words.	Read
`get_task`	Everything about one task, with comments and history.	Read
`workspace_overview`	The Dashboard: overdue, due soon, recent activity.	Read
`board_report`	Totals, completion, workload.	Read
`list_notes`	Notes on a board.	Read
`get_note`	One note with its body.	Read
`create_task`	Create a task.	Write
`bulk_create_tasks`	Create up to 50 tasks at once.	Write
`update_task`	Change title, dates, priority, assignee, tags...	Write
`move_task`	Move to another status or project.	Write
`add_comment`	Comment on a task.	Write
`add_subtask`	Add or link a subtask.	Write
`add_dependency`	This task waits on another.	Write
`remove_dependency`	Remove a dependency.	Write
`archive_task`	Archive a task.	Write
`unarchive_task`	Restore an archived task.	Write
`create_note`	Create a note, optionally linked to a task.	Write
`update_note`	Change a note.	Write

A read token only sees the read tools. Tool calls are limited to 120 a minute per token.

Tip Keep one agent per tool (for example Claude Code and Cursor), so the task history shows which one made each change.

Was this page helpful?

PreviousAI agents and tokens NextSign in