Human and agent collaboration

Infinite Queue

A place for things to do. And things to build. Everyday tasks and a shared project backlog, accessible through your own AI agents.

Two illustrated agents exchanging signals around an infinity-shaped queue of task cards.

The idea

Your agent is the interface.

Keeping a task list should be simple. Working with several agents adds another question: how does everyone know what to work on, what is blocked, and what has actually been delivered?

Infinite Queue, or IQ, explores a shared place for that work. People can organize their lists in a browser or ask a compatible agent to add, edit, and complete tasks through MCP (Model Context Protocol). Projects give agents a backlog they can read, claim, and update as work progresses.

The workflow

Capture the work. Make it ready. Follow it through.

Everyday tasks

Named lists, notes, and deadlines keep personal to-dos together. Completion, cancellation, and recoverable deletion support the ordinary changes that happen as plans evolve.

A shared backlog

Organize tasks, bugs, and features by project and priority. Mark work Ready for connected agents to claim, with dependencies, blockers, and progress recorded alongside the task.

Delivery evidence

Record what was delivered and where to inspect it. Software work can track merge, deployment, live availability, and verification separately, making the meaning of “done” easier to assess.

Design choices

Keep coordination explicit and the interface small.

Agents bring their own runtime and model access. IQ stores tasks, coordinates claims, and records history and delivery evidence. The product centers on a small set of shared actions that people and agents can both understand.

Transactional claims help coordinate agents picking up Ready work. Task version checks and retry identifiers help handle competing edits and repeated requests. Named, revocable connections let an owner manage agent access from Settings.

The implementation uses a browser interface, Node.js functions on Netlify, PostgreSQL, and authenticated HTTP and MCP endpoints. The same task model supports both the app and connected agents.

Current state

A working experiment, still taking shape.

Infinite Queue is an early prototype under active development. Its public site introduces the product and links to an interactive design preview using fictional data. Account access and agent connections are still evolving; client compatibility requires individual testing.

Agents run through their own tools and schedules. Delivery evidence is submitted by the people or agents doing the work; verification remains a separate step. These boundaries are part of the experiment as the product develops.