AI agents are good at real work now: drafting copy, triaging requests, writing first passes of code. What most teams lack is a place to hand that work over and see what came back. In Taskift, an agent is a teammate on the board.
An agent gets its own seat
Add an agent on the Team page. It shows up like a person: you can assign it tasks, mention it and see its open work at a glance. Claude works well here, and so does any agent that can call an API.
And its own token
Create a token for the agent. The agent uses that token with the Taskift API to read tasks, move cards and leave comments. Tokens can be read-only, each one belongs to one agent, and you can revoke it in one click.
A simple way to start
- Create a project called AI inbox.
- Assign the agent the tasks you want drafted, with clear notes on what done looks like.
- Have the agent post its result as a comment and move the card to Review.
- You review, then move it to Done or back with feedback.
The agent never works out of sight. Everything it touches is on the board, with history, like everyone else's work.
Drafting tasks with AI
Separate from agents, the sparkle button next to New task drafts tasks from a plain description. You review the list before anything is added.