Can I run several Claude Code sessions on one project at once?

Yes, on Pro (up to 5 concurrent sessions). Open several Claude Code windows on the same project — they share one queue and it divides itself: each session leases the next unclaimed task via get_next_task, so they work the top tasks in parallel without two agents grabbing the same one. One session works the list in sequence; five work five tasks at once. Each session also auto-scopes to the project from its working directory.

Is TaskPrio free for Claude Code?

Yes — the Free plan includes 1 MCP session and 50 agent completions per day, enough for everyday Claude Code use. Pro ($10/month) raises that to 5 concurrent sessions and unlimited completions.

The task manager for Claude Code

Q: How do I give Claude Code a task queue?

Install TaskPrio's MCP server with one line — curl -fsSL https://promptprio.com/install.sh | sh — and restart Claude Code. Claude Code then has tools like get_next_task and complete_task, so it can pull the highest-priority task from your queue, work it, and report a result on a loop.

Q: Can Claude Code work tasks on autopilot?

Yes. TaskPrio keeps one ranked queue; Claude Code calls get_next_task to receive the top unblocked task with context, completes it, then loops. It only stops when the queue returns no task, hits a blocker (flag_needs_human), or you stop it.

Give Claude Code a real task queue. TaskPrio is MCP-native — Claude Code pulls the highest-priority task with get_next_task, works it, reports a result, and grabs the next one. You set the order; it runs the list on autopilot — and you watch it work in a live cockpit.

Open the board — free →

Connect in one line

curl -fsSL https://promptprio.com/install.sh | sh

This downloads the bundled MCP server and merges a promptprio entry into your Claude Code MCP config. Restart Claude Code and the tools are live. (Prefer to do it by hand or self-host? The docs cover both.)

The autopilot loop

Once connected, point Claude Code at your queue and it runs top to bottom — auto-scoping to the project your terminal is in (it reads the working directory, no manual setup):

1. get_next_task     → top unblocked task + merged project context
2. (Claude Code does the work in your repo)
3. complete_task     → short result, logged as an audit trail
4. repeat            → until the queue returns no task

Hit a blocker? Claude Code calls flag_needs_human and moves on, so one stuck task never stalls the whole run.

Why a queue beats a prompt dump

One priority order — Claude Code always knows the single next thing, even across 1,000+ tasks and many projects.
Context comes with the task — each task carries its project's merged instructions, so you stop re-pasting setup into chat.
Parallelise across sessions (Pro) — open up to 5 Claude Code windows on the same project and the queue divides itself: each session leases the next unclaimed task, so five sessions work the top five in parallel — no partitioning, no two agents on one task.
You stay in control — capture tasks in the Inbox, rank what matters, and the agent only works what you've ranked.

Frequently asked questions

How do I give Claude Code a task queue?

Install the MCP server with one line and restart Claude Code. It then has get_next_task, complete_task and friends, so it can pull from and complete tasks in your TaskPrio queue.

Can Claude Code work tasks on autopilot?

Yes — it loops get_next_task → work → complete_task until the queue is empty, a task is blocked, or you stop it.

Can I run several Claude Code sessions on one project?

Yes (on Pro) — open up to 5 Claude Code windows on the same project. They share one queue and it divides itself: each session leases the next unclaimed task, so they work the top tasks in parallel without colliding. One session works the list in sequence; five work five at once.

Is it free?

Yes. Free includes 1 MCP session and 50 completions/day; Pro ($10/mo) gives 5 sessions and unlimited completions.

Start free — no card → Using Cursor? →

Run it on autopilot. Paste our maintained autopilot prompt into your agent — it adds hard safety gates (money / credentials / publishing stay manual), an empty-queue stop, and auto-scopes to your project.

view ↗