Show HN: TaskPeace – a task queue my AI coding agents pull work from over MCP

TaskPeace — Task Manager for AI Coding Agents (MCP-Native)

The new bottleneck isn't intelligence. It's collaborating with it.

Mission control for your AI coding agents.

One ranked queue your agents pull work from — and a live cockpit you watch them in. Give Claude Code, Cursor or any MCP agent a backlog; each calls get_next_task, works the top task, and reports back — top to bottom, on autopilot. You decide what matters and see exactly what every agent is doing. You stay in control.

1 · CaptureDrop in tasks, prompts, links, files — humans and agents share one list.

2 · Rank onceDrag into a single global priority. No statuses to babysit.

3 · Agent works itIt calls get_next_task, does it, marks it done — top to bottom.

Works with Claude Code · Cursor · ChatGPT · REST API · built-in MCP

The board below is a live demo (saved in your browser). Sign in to sync it with your agent. Free for humans · Pro a flat $10/mo — no per-agent fees, no extra tokens · Learn more

Agents shipping right now

Connect your agent in one line

TaskPeace is MCP-native (Model Context Protocol — the standard way agents connect to tools): install the server once and your agent works the queue on autopilot. No glue code, no orchestration to maintain.

1 · Install · Claude Code / Cursor

curl -fsSL https://taskpeace.com/install.sh | bash

2 · Your agent runs the loop

get_next_task # top task + context → do the work → complete_task # logs a result → repeat # until the queue is empty

Setup guides: Claude Code · Claude Desktop · Cursor · Cline · Goose · Warp · Continue · Windsurf · Zed · Codex CLI · Gemini CLI · ChatGPT · or the REST API.

MIT-licensed · self-hostable · free to start — source on GitLab

What is TaskPeace?

TaskPeace is mission control for AI coding agents — one ranked queue your agents pull work from, and a live cockpit you watch them in. You keep one global priority order across Claude Code, Cursor, ChatGPT and every project; each agent calls get_next_task over MCP or REST, receives the highest-ranked task with full context, works it, and reports back. It's the layer above any single tool's session: you decide what matters, your agents run it top to bottom on autopilot, and you see exactly what each one is doing.

Why TaskPeace exists

The new bottleneck in software isn't intelligence — frontier models are already capable enough. It's collaboration: getting an agent to work on the right thing, in the right order, with the right context, then report back so you can trust it ran unattended. TaskPeace makes that concrete — one shared priority queue, above every agent and tool, where you decide what matters and your agents execute it top to bottom. As single-tool coordination becomes commoditized, the durable layer is the one that spans all your agents and keeps a human in the loop. You own priority; the agents own execution.

How is TaskPeace different from Notion, Linear, or Trello?

TaskPeace has exactly one ranked queue across every project. Notion, Linear, and Trello let you have many priorities — three P0 labels, four Today columns, two Urgent tags — which collapses into "I don't know what to do next" the moment you sit down. TaskPeace's one queue answers "what's next" unambiguously, for you and for your agent. See the full comparison for AI-agent work →

How much does TaskPeace cost?

Free for humans with generous limits: 200 active tasks, 5 projects, 1 MCP session, 50 agent completions per day. Pro at $10/month unlocks unlimited everything — tasks, projects, agent completions — plus 5 concurrent MCP sessions, 4 MB attachments, team boards, and webhooks. See /pricing for the exact limits table.

Does TaskPeace work with Claude Code, Cursor, and ChatGPT?

Yes — anything that speaks Model Context Protocol (MCP) or REST. Claude Code and Cursor consume the bundled MCP server via a one-line installer at taskpeace.com/install.sh. ChatGPT and other LLMs hit the REST API directly using a Bearer token from your account. The full agent-discoverable manifest lives at /.well-known/agent-tasks.json. Setup guides: task manager for Claude Code, for Cursor, and how the MCP task queue works.

Where is my data stored?

Cloud-mode boards live in Upstash KV (encrypted at rest, region-pinned per account). Local-mode boards stay in your browser's localStorage and never touch any server — you can use TaskPeace fully without an account. Switch between modes at any time without losing data. See /privacy.

Agent Memory Compiler · AGENTS.md · AGENTS.md vs CLAUDE.md · CLAUDE.md · GEMINI.md · Cline rules · Copilot instructions · Cursor rules · AI agent task manager · MCP task manager · Claude Code · Cursor · ChatGPT · Compare · vs Agent Teams · vs Linear · vs Jira · vs Asana · vs Todoist · vs Notion · vs GitHub Issues · vs Shrimp · Guides · Agentic workflows · Docs · Methodology · Pricing · API reference · About

Columns

↑↓ nav · Enter select · Esc close · ⌘ K open

Your queue as an AI prompt

Paste this into ChatGPT, Claude, or any LLM chat — it's your ranked queue, formatted as a prompt that asks the AI to work the top task first. Your agent gets the same content via the get_next_task MCP tool / /api/next.

Account

· plan

Agent token

Install MCP — paste the copied one-liner into your terminal. It verifies the token, downloads the bundled MCP server to ~/.local/share/promptprio/, and merges a promptprio entry into ~/.claude/mcp.json. Restart Claude Code after.

Shell setup writes export PROMPTPRIO_API_TOKEN=... to ~/.zshenv and removes any prior line.

MCP config is the manual JSON path — only if you'd rather edit ~/.claude/mcp.json yourself.

Never paste your token into chat — only your own terminal or your own config file.

Keyboard shortcuts

Anywhere on the board (skipped while typing in a field).

jNext card · move selection down kPrevious card · move selection up Enter / eEdit selected task xComplete selected task rRank selected Inbox task → top of queue p▶ Play selected task in your preferred LLM (Claude Code / Cursor / web). Shift P forces the launcher picker. cFocus the quick-capture input (sidebar) nNew task — full dialog (body, schedule, attachments, properties) /Focus the search bar 1 … 9Jump to column N (scroll into view + expand if collapsed) fFocus / un-focus the current column (full-screen) tToggle light / dark theme ⌘ KCommand palette (jump to task / project / org / run command) ⌘ \Toggle sidebar ?Show this help EscClear search · exit focus · close dialogs ⌘ ZUndo the last change

Quick capture syntax: #project · @agent / @me / @any · !top / !bottom / !inbox · +OptionLabel (matches any single-select option, e.g. +Doing)

Set up TaskPeace — from connected to fully autonomous

Everything worth installing & connecting, in order. Stop at any tier — tier 1 is all you need to start.

1 Connect your agent required

One line gives Claude Code / Cursor / ChatGPT the queue tools (get_next_task, complete_task, …). Grab your token first: sidebar foot → Connect your agent → Copy token, then:

PROMPTPRIO_API_TOKEN=pp_YOUR_TOKEN bash <(curl -fsSL https://taskpeace.com/install.sh)

Restart your agent, then verify: claude mcp list → promptprio ✓ Connected. Per-tool guides: Claude Code · Cursor · Cline · Goose · Warp · Continue · Codex CLI · Gemini CLI · ChatGPT · Windsurf · Zed.

2 Work the queue on autopilot

The loop — six prompts, one operating cycle. plan fills the queue with the right work → autopilot / never-stop clears it → continue resumes it if a session ever stops → capture folds any conversation back in → every session quietly trains TaskPeace itself → stop winds down clean. Paste the one you need; together they run the whole operation, with you in the loop only where your judgment is irreplaceable.

Start by filling the queue with the right work — paste the plan prompt on a project (or right after you set a goal). It turns the goal + live metrics into a ranked, well-formed backlog, routed to the right worker, so autopilot then has work genuinely worth clearing:

view ↗

Or skip the goal-and-metrics route — just describe what you want and TaskPeace wraps it into the plan prompt, so your agent turns it straight into ranked, well-formed tasks. You never write the prompt:

Tell your agent to work the queue — it loops get_next_task → do it → complete_task and auto-scopes to the project your terminal is in (from the working directory). No need to say which project. Blocked tasks become a human flag, so one stuck item never stalls the run.

For the most robust run, paste our autopilot prompt — it adds hard safety gates (money / credentials / publishing stay manual), self-configures the working directory, and stops on its own when the queue is drained:

view ↗

Want it to run forever until you say stop? Paste the never-stop prompt instead — same autopilot, but on an empty queue it waits (≈4 min) and re-polls rather than ending, so the moment you add a task or one comes due it's already on it. The wait keeps it cheap — it never busy-polls. Only stop / pause / halt ends it:

view ↗

Session stopped — crash, context limit, closed tab, rate-limit, lost connection? Paste the continue prompt to pick up cleanly: it re-orients first (recovers any half-done task, never re-does what already shipped, makes the board true), then resumes the loop in whichever mode it was running. Resume from anywhere — nothing lost, nothing double-done:

view ↗

Already in an autopilot run? You don't have to re-paste — just send q (or continue) on its own and it picks the loop back up (it leads with ▶ q · TaskPeace autopilot — resuming… so you see it caught). Only stop / pause / halt ends a run. The installer also teaches your agent the q key globally (a marked block in your CLAUDE.md), so it works from a fresh session too — every session knows it.

Done for now? Paste the stop prompt to end the session cleanly — it lands the current task (done-with-proof, or a resume note), syncs the board true, banks one learning, and leaves you a short report:

view ↗

Had a valuable conversation that isn't an autopilot run — a long Claude Code session, a claude.ai chat full of decisions and ideas? Paste the capture prompt at the end and it folds everything worth keeping — follow-ups, decisions, learnings — into your queue as ranked, resumable tasks. Conversations feed the queue; the autopilot prompt works it. Nothing good is lost:

view ↗

3 Run unattended / overnight optional

Keep a session alive after you close the window, and keep the Mac awake. Install once (macOS):

brew install tmux && brew install --cask ghostty

Then launch a persistent, awake session — detach with Ctrl-b d, reattach with tmux attach -t ace:

tmux new -s ace caffeinate -dimsu claude

For hands-off runs, let the agent act without per-step prompts — set "defaultMode": "bypassPermissions" in your agent's settings, and keep a deny-list for destructive commands. Money, credentials, new accounts & irreversible deletes should always stay manual.

4 Parallelise Pro

Open up to 5 sessions on the same project — the queue divides itself: each session leases the next unclaimed task, so five sessions work the top five in parallel without collisions. The board shows which session holds each task. One session works the list in sequence; five work five at once.

5 Public profile optional

Turn your shipped work into a shareable credential at taskpeace.com/@you — tasks shipped, median cycle time, and which agents you run. Off by default. When public, only aggregate stats show — never task titles or contents.

taskpeace.com/@

make public

Full reference: docs · MCP task manager · REST API · agent manifest.

Analytics

Completions, cycle time, and who's shipping. Time-to-done counts from a task's creation to its completion.