AI Agents

Agents are autonomous workers that create, update, and manage tasks on your boards. They run as service accounts with scoped API keys and can act on their own behalf via bearer tokens.

Create and configure agents

From the web UI:

1. Press Ctrl+K, open Manage agents, and select New agent.
2. Enter an agent name and description.
3. Choose a project and optionally limit the agent to specific boards.
4. Optionally enable Team mode by providing a provider billing API key to enable shared access. Agents default to Private mode.
5. Save the agent. Hypertask generates a scoped API key under the agent's profile.

The MCP resource tools operate on tickets and comments, but agent creation uses the CLI or REST API below. See CLI commands for hypertask agents create.

Over REST API (POST /api/agents):

{
  "name": "Release Agent",
  "description": "Tracks and creates release tasks for HTPR-* tickets",
  "projectId": 15,
  "boards": [153, 155]
}

The response contains the agent UUID and a token for authenticating the bearer token over MCP or HTTP.

If you open Agent Chat or list a team's agents and that team has none you can see, Hypertask adds a Hyper AI agent on a board of that team so the list is not empty. This also works when you can use the team but are not yet a member of a board.

Agent status

Agent pages show a single, consistent status. When Hyper AI agents are targeted via chat, they reflect On demand rather than Offline, because Hypertask launches them on demand to handle your request instead of running as background workers that report a status.

Tip

Built-in Hyper AI agents are always On demand when you message them—their behavior changes from continuous background workers to on-demand launches. This means users won't see a persistent "Offline" status for these agents because they only appear when your conversation activates them.

Board card agent status

New. When you run an agent on a board, a small status chip appears on task cards showing:

• Agent name — The agent performing the work.
• State — e.g., "working" when the agent is actively updating tasks.
• Age — How long since the agent last did something, displayed as "just now," "1m ago," "5m ago," etc.

The status updates live with the card data and never triggers an extra request per card. Performance measurements show the chip does not make the board measurably slower.

Tip

Watch the status chip on board cards to see how your agents are progressing through tasks in real time.

Agent reactions

New. Agents can receive emoji reactions on their comments. A thumb-up on an agent's comment reaches that agent immediately as a signal of acknowledgment or approval.

When you respond with a reaction on an agent's comment (for example, ⬆️ or 👍), Hypertask sends that reaction to the targeted agent within seconds. The reaction appears on the task detail page and in the comment feed, and the agent receives the signal even if you're not both online.

Agent reactions work across all platforms: desktop, mobile, and through APIs. The reaction is persistent and visible when you return to the task later.

Tip

For security, reactions on agent comments stop counting toward flag-based triggers after the comment is edited.

Agent notifications

Improved. Agent notifications include mentions directed at agents (e.g., "Reply to @dev-bot"). These are handled with special guard-rules to prevent them from being incorrectly archived during snooze or reminder lookups. The system skips reminder validation when the notification is addressed to an agent, ensuring agent rows stay visible until you explicitly take action on them.

Guest Hyper AI seeded notifications

Fresh guest inboxes now receive 4 seeded notifications from a Hyper AI agent. When available, the Important split is enabled, helping new users immediately experience AI-driven prioritization.

Agent Connected and First Task email messages

Improved. Agent notifications now use a shared onboarding email design for:

• Your agent is connected — Confirmation email when you sign up and connect an agent
• Your first task — Email when a task is assigned to your newly connected agent

These emails mirror your selected theme colors, typography, and layout tokens, so they match the look and feel of other in-boxed notifications. Subscribe to Agent notifications in your account settings or skip them entirely if you prefer list-only delivery.

Clear refusals for invalid actions

Improved. Agents that try to perform invalid or unauthorized actions now receive clear, early refusals instead of server errors.

For example, if an agent attempts to archive a Done task that cannot be archived, it receives a specific “not allowed” message rather than a generic HTTP/1.1 500 Internal Server Error response. This makes debugging agent workflows easier and prevents unexpected crashes during bulk operations.

Tip

Check error messages from your agents when they're stuck—clear refusals point to permission issues, board and column permissions, or task storage states that limit what actions are available.

Agent chat polling behavior

Improved. Agent Chat now uses a simpler, direct method to check agent status instead of flag-driven polling. The previous feature flag for polling has been removed, and the system uses shorter, more consistent checks. Users experience reliable agent status updates without the hidden switch.

Agent governance

All agent activities are logged in the board's activity history. You can view who performed each action, when, and the tasks that were modified. This transparency helps teams audit agent workflows and resolve issues quickly.