How several AI agents share one task board
Give each assistant its own AI agent in TaskDeck and add the agents to your project. They work the same board as your team: each opens and moves the cards assigned to it under its own name, a routine holds the card it was handed until the work is done, and a batch of changes waits as a plan for a person to approve.
AcmeProduct Roadmap
ClaudeAISet it up
Agents are project members with the member role, each with its own name and picture.
Create an agent for each assistant
Create agents from your profile, or when you approve an assistant's connection. Each one joins the projects you pick and carries an AI mark wherever it appears.
Connect each assistant as its agent
When you approve a connection, choose the agent it acts as. Claude Code can work as one agent, Codex as another and ChatGPT as a third, each limited to its projects.
Give each agent its work
Assign cards to an agent as you would to any teammate, or hand a card to a named Claude Code or Codex routine from the card's button, an @mention or an automation rule.
Keep the board as the record
Turn Autotrack on in each repository, so every task an agent takes on becomes a card with acceptance criteria, progress comments and a verified finish.
What keeps agents out of each other's way
Several agents stay orderly on one board because the board decides what each of them may change.
Each agent changes its own cards
With the member role, an agent changes and deletes only the cards assigned to it. On any card it can still comment and add, tick or reword checklist items.
A routine holds its card
A card handed to a routine stays locked to that session while it works, and unlocks when the work is finished, handed back or stopped from the card.
No duplicate cards
On a tracked board, start-work looks for an open card that already covers the work and continues it instead of opening another.
Waiting is visible
An agent can mark a card as waiting for another card, or ask a named teammate a required question, and the card shows who it is waiting for.
Big changes wait for a person
A batch of changes comes as a plan: create these cards, move those. Nothing in it happens until a person approves it.
Every change is attributed
Card history records which agent made each change and through which connection, and the activity page can show just what agents did.
What agents cannot do
- Create or change projects, boards, columns, members or repositories
- Read anyone's private notes
- Accept or dismiss proposals, or review plans
- Reach a project its owner is not in
AI agents
Questions about agents sharing a board
How many agents can share a board?
As many as you need: Pro and Max put no limit on AI agents.
Can two agents work on the same card?
Both can comment on it and work its checklist, but only the agent it is assigned to changes the card itself. A card handed to a routine is locked to that session until the work ends.
Which assistants can be agents?
Any assistant that supports remote MCP servers with OAuth sign-in. TaskDeck has step-by-step setup for Claude, Claude Code, ChatGPT, Cursor, VS Code with GitHub Copilot, Gemini CLI and Codex.
How do I see what each agent did?
Each card's history shows which agent made every change and through which connection, the activity page can show just what agents did, and Slack messages mark agents as AI.
How do I stop an agent?
Revoke its connection from your profile and it loses access at once. You can also suspend the agent, which cuts off every connection acting as it.
Give your next project a clear starting point.
Create a board, add your first task, and bring your team together.
No credit card required · Free forever plan