AI Agent Task Manager

AI Agent Task Manager gives developers a board for queuing and monitoring parallel coding-agent sessions. Agents pull tasks, report results, and flag when they need a human decision.

One of 672 tools in AI Agents

Visit website Compare alternatives Own this tool? Claim it

Who it's for

Developers running several Claude Code, Codex, or Cursor sessions who need one view of each session's task and status.
Engineering teams splitting feature, API, and test work across independent coding-agent queues.
Developers using custom scripts or agent frameworks that can make HTTP requests and need a REST task queue.

Not the right fit if…

Teams that need agents to communicate through a shared chat relay: the page describes the board as a status signal, not a chat relay.
People who need more than 24 concurrent workstreams on one board: the page gives 24 as the soft cap.

How it fits your workflow

Create one workstream per agent session, then add ordered task cards with instructions. The board can also be used manually without connecting an agent.

To automate a stream, generate an API key and copy the bootstrap prompt into a fresh supported agent session. The agent pulls the top task over HTTPS, works on it, marks it complete with a short result, and requests the next task.

When an agent needs a decision, it flags attention. The board marks the workstream and sends a browser notification; the user responds in the agent's terminal, and the agent resumes.

Pricing

Free: The board is free. Anonymous use supports manual board use in the browser; signing in with Google is required to connect agents and sync across devices.

Free

· No signup required to try the board manually; anonymous boards save in the browser using IndexedDB.
· Sign in with Google to connect agents and sync across devices.
· Up to 24 workstreams per board, described as a soft cap.
· Unlimited tasks per stream and unlimited completed tasks.

Prices checked on Oct 21, 2026 from the vendor's site. They can change; confirm before you buy.

Key features

Separate workstreams and task queues: Create a workstream for each agent session. Each has an ordered queue, and agents work on the top queued task before moving to the next.
Agent bootstrap prompt: The Connect an agent flow generates a prompt to paste into a Claude Code, Codex, or Cursor session, explaining how to pull tasks and report progress.
REST agent protocol: Agents can claim tasks, flag attention, resume after a human response, and complete tasks with a short result summary using the board's REST endpoints.
Needs-you status and notifications: An agent can raise an attention flag when it needs a decision. The relevant column lights up and the browser can notify the user; Web Push is available when subscribed.
Long-poll task retrieval: The agent next-task endpoint supports a wait parameter of up to 30 minutes, allowing a connection to wait for a task rather than repeatedly polling.
Manual task-board use: The board can be used without connecting agents, including as a place to track what each session is doing and update tasks by hand.

Works with

Claude Code
OpenAI Codex
Cursor
HTTP-capable agents and scripts
Google sign-in

Limitations to know

The free board has a soft cap of 24 concurrent workstreams per board.
Connecting autonomous agents and syncing across devices require signing in with Google; anonymous use is for browser-local boards.
On touch devices, drag-and-drop can be fiddly; the page recommends using the task edit modal to move cards across columns.
Running and needs-you tasks are locked and cannot be moved between workstreams while in those states.

Getting started

To connect an agent, sign in with Google and use an agent session or script that can make HTTP requests. Anonymous use is available for manual board use.

Open the board and create a workstream for each agent session.
Add tasks with a title and detailed instructions.
Select Connect an agent, generate a personal API key, and copy the bootstrap prompt.
Paste the prompt into a fresh Claude Code, Codex, or Cursor session and monitor the board for progress or attention flags.

Frequently asked questions

Is it free, and do I need to sign up?

The board is free and can be used anonymously without signup; anonymous boards save to the browser using IndexedDB. Sign in with Google to connect autonomous agents and sync across devices.

How do agents connect to the board?

Select Connect an agent on a workstream, generate a personal API key, and paste the generated bootstrap prompt into a fresh Claude Code, Codex, or Cursor session. The prompt tells the agent how to pull tasks, complete them, and repeat through the HTTPS API.

What happens when an agent needs a decision?

The agent flags the task, the workstream indicates that it needs you, and the browser can send a notification. You respond directly at the agent's terminal, after which the agent can clear the flag and resume.

Can the board be used manually?

Yes. The board can be used without agent automation to track sessions and update tasks by hand.

Can it work with agents other than Claude Code, Codex, and Cursor?

Yes. The board's protocol uses HTTP, so any agent framework or script that can make a curl request can drive it.

Does an agent keep working if the browser tab is closed?

Yes. The agent runs in its own terminal session on the user's machine; the browser tab is a status surface. The page also says Web Push can notify the OS when an agent flags attention, including with a closed tab when enabled.

How many workstreams can a board have?

The page states that a board can run up to 24 workstreams concurrently, a soft cap intended to keep the UI usable. It also states there are unlimited tasks per stream and unlimited tasks completed over the lifetime of the board.

Is the API key scoped to the user's board?

The page says the key is a long random token, stored as a SHA-256 hash, and scoped to read and write only the user's own workstreams and tasks. It can be regenerated to invalidate the previous key.