precision_manufacturingBuilt for Autonomous Operation

A Task Queue Your AIAgents Can Actually Use

Give your coding agents a proper task queue. Agents fetch the next task, complete it, log progress, open a PR. You review and steer.

dashboard
Agent Queue DashboardMonitoring 3 active agents
smart_toy
Agent-1
Working
Task #47: OAuth Login
Priority: High
Progress:67%
smart_toy
Agent-2
syncFetching next
check_circleCompleted Task #38
Duration: 42 min
Logged 0.7h effort
smart_toy
Agent-3
Working
Task #52: API Tests
Priority: Medium
Progress:23%
Recent Activity:
03:45:12> Agent-2 completed Task #38
03:45:15> Agent-2 logged 0.7h effort
03:45:18> Agent-2 fetching next task...
03:34:22> Agent-1 progress: 67%
03:21:10> Agent-3 progress: 23%
24/7Autonomous
3Active Agents

Getting Started

Total beginner? You only need three things.

1

Install the runner

Add the mcptask_runner gem to your Rails project, run the installer, and you are connected in about two minutes.

See the full install walkthrough
2

Get your MCPTASK_TOKEN

In mcptask.online, open Settings → API Access and generate a personal access token. The installer will prompt you to paste it during setup.

Open mcptask.online
3

Tell the AI which project to work on

Add a short mcptask.online section to your project's CLAUDE.md with your project_relative_id and account_code. Without it, the agent has no idea where to read or write tasks.

Jargon, in plain English

MCP (Model Context Protocol)
The open standard AI clients use to talk to external tools. Think of it as a USB cable between Claude Code and mcptask.online.
Runner
The mcptask_runner gem + a daily LaunchAgent (macOS) that fetches one task, lets your AI work it, opens a PR, and repeats.
Agent
Your connected AI client — Claude Code, Claude Desktop, or any MCP-compatible assistant. Agents are free on every plan.
Quota
The daily cap on how many tasks the runner will work autonomously. Stops the loop, no surprises.
LaunchAgent
macOS-only scheduler. The installer registers a LaunchAgent that fires the runner on weekdays at 08:00. On Linux/Windows, use cron, systemd, or Task Scheduler.

The Autonomous Agent Challenge

1

No Standardized Task Queue

You want to run AI agents while you sleep, working through your backlog. But there's no standardized way to: Give agents a queue of tasks, Let agents pick the next priority, Track what agents are working on, Know when agents finish.
warning

Current Workaround

Custom scripts, JSON files, manual assignment

error_outline

Problem

Fragile, no visibility, doesn't scale

2

Black Box Operation

Agent runs for hours. Morning arrives. What happened? Which tasks were attempted? What was completed? What failed? Why did it stop?
info

Current State

Chat logs, scattered files, no structured data

error_outline

Impact

No confidence in agent work

3

No Human Oversight

You want agents to work autonomously, but you also need: A way to stop when things go wrong, Review of every PR before merge, A clear audit trail of what each agent did.
info

Current State

All or nothing - full auto or manual

psychology

Need

PRs, effort logs, and a quota cap

4

Scaling Challenges

One agent works. Three agents conflict. Task duplication, race conditions, no coordination.
error_outline

Problems

Same task picked by multiple agents, No work distribution logic, No capacity management

mcptask.online as Your Agent's Task Queue

The Agent Workflow

1

Agent calls get_next_task() via MCP

arrow_downward
2

mcptask.online returns highest priority task

arrow_downward
3

Agent works on task (coding, testing, etc.)

arrow_downward
4

Agent calls log_effort() with progress

arrow_downward
5

Agent calls complete_task() when done

arrow_downward
6

Loop: Agent calls get_next_task() again

check

Task locking prevents duplicate work

check

Priority-based task selection

check

Automatic effort logging

check

Clear completion tracking

check

Continuous operation loop

Feature Highlights

queue

Smart Task Queue

Agent calls get_next_task() and receives:

  • check_circleHighest priority available task
  • check_circleAutomatically locked (no other agent can pick it)
  • check_circleFull context (description, acceptance criteria, related tasks)
  • check_circleTime estimate (if available)

Priority Factors

Priority levelUrgent > High > Normal > Low
Due datesooner = higher priority
Task typebugs before features, configurable
Ageolder tasks surface

Scoping Options

  • arrow_rightLimit to specific projects
  • arrow_rightLimit to specific task types
  • arrow_rightExclude certain labels
  • arrow_rightTime-based restrictions
history

Full Activity Logging

Every Action Recorded:

  • terminalTask fetched by Agent-1 at 02:34:15
  • terminalAgent-1 logged 45 min of work at 03:19:22
  • terminalAgent-1 completed task at 03:45:08
  • terminalAgent-1 fetched next task at 03:45:12

Morning Dashboard Shows:

  • check_circleTasks completed overnight
  • check_circleTotal hours logged
  • check_circleSuccess/failure rate
  • check_circleAny blocked tasks
  • check_circleAgent activity timeline
visibility

Human Oversight Controls

Safety Rails:

Control Types

Task TypesAgents can only work on approved task types
Approval GatesCertain tasks require human approval before agent starts
Daily QuotaCap on how many tasks the runner will work autonomously per day
LaunchAgent (macOS)Schedule the runner to fire on weekdays at 08:00 via a macOS LaunchAgent
Stop the RunnerThe runner is just a rake task — Ctrl-C, kill the process, or disable the LaunchAgent

notification_importantAlerts

  • arrow_rightPR opened (review in GitHub)
  • arrow_rightAgent encountered blocker (visible in effort log)
  • arrow_rightAgent failed task (visible in effort log)
  • arrow_rightReal-time webhook notifications — coming soon
groups

Multi-Agent Coordination

How It Works:

  • check_circleEach agent has unique API key
  • check_circleTask locking prevents conflicts
  • check_circleWork distributed automatically
  • check_circleNo race conditions

Agent Fleet Management

  • check_circleSee all active agents
  • check_circleMonitor individual progress
  • check_circlePause/resume specific agents
  • check_circleReassign work between agents

Scaling

Run 1 agent or 100. Same interface. Same pricing (agents are free).

Agent Configuration Examples

Illustrative agent concepts. The real setup is a gem + a .mcp.json + a daily LaunchAgent — see the install guide.

info

Note: these YAML blocks are illustrative. mcptask.online does not parse them — your agent does. The actual configuration is the .mcp.json below.

The real, copy-pasteable config

What your agent actually loads — one MCP server entry, one env var, one config file:

code.mcp.json
{
  "mcpServers": {
    "mcptask-online": {
      "type": "sse",
      "url": "https://mcptask.online/mcp/sse",
      "headers": { "Authorization": "Bearer ${MCPTASK_TOKEN}" }
    }
  }
}
bug_report

Overnight Bug Fixer

Agent works overnight on bugs only, stops on error.

codeYAML Configuration
agent_name: "BugFixer-1"
scope:
  projects: ["main-app"]
  task_types: ["bug", "hotfix"]
  priority_min: "normal"
limits:
  max_tasks_per_session: 20
  working_hours: "22:00-06:00"
behavior:
  on_complete: fetch_next
  on_error: pause_and_alert
  log_frequency: per_task
group_work

Feature Implementation Team

Specialized agents for different work types.

codeYAML Configuration
agents:
  - name: "Frontend-Agent"
    scope:
      labels: ["frontend", "ui"]
  - name: "Backend-Agent"
    scope:
      labels: ["backend", "api"]
  - name: "Test-Agent"
    scope:
      task_types: ["test", "qa"]
    behavior:
      requires_approval: true
verified_user

Conservative Autonomous Mode

Agent completes task, pauses for human review, then continues.

codeYAML Configuration
agent_name: "Careful-Agent"
scope:
  projects: ["low-risk-project"]
limits:
  max_tasks_per_day: 5
  require_approval_for: ["feature", "refactor"]
behavior:
  on_complete: pause_for_review
  alert_on_every_completion: true

Safety & Control

Autonomy with Oversight

We believe autonomous agents should be powerful AND controllable. mcptask.online provides:

lock

Scoped Access

  • check_circleAgents only see authorized projects
  • check_circleAgents only work on approved task types
  • check_circleGranular permission control
receipt_long

Audit Everything

  • check_circleEvery agent action logged
  • check_circleFull history exportable
  • check_circleClear attribution
emergency

Stop the Runner

  • check_circleCtrl-C the rake task to stop in-flight work
  • check_circleDisable the LaunchAgent to stop future runs
  • check_circleRevoke the agent's MCPTASK_TOKEN to lock it out
how_to_reg

Human Gates

  • check_circleEvery change goes through a pull request
  • check_circleCI must be green before auto-merge
  • check_circleNo direct pushes to main
notification_important

Alerts

  • check_circlePR opened (visible in GitHub)
  • check_circleErrors in effort log
  • check_circleReal-time webhook notifications — coming soon

Monitoring Dashboard

What You See

dashboard

Agent Status Panel

  • arrow_rightActive agents and current task
  • arrow_rightIdle agents
  • arrow_rightPaused agents
  • arrow_rightError states
timeline

Activity Timeline

  • arrow_rightTask started by Agent-X
  • arrow_rightTask completed by Agent-Y
  • arrow_rightEffort logged
  • arrow_rightErrors encountered
analytics

Metrics

  • arrow_rightTasks completed (24h / 7d / 30d)
  • arrow_rightHours logged by agents
  • arrow_rightSuccess rate
  • arrow_rightAverage task duration
  • arrow_rightThroughput trends
warning

Alerts

  • arrow_rightCritical errors
  • arrow_rightBlocked agents
  • arrow_rightUnusual patterns

How Teams Use It

format_quote
We run our coding agents through the queue. Every morning we review what they completed. We get more done without adding headcount.
I

Illustrative scenario

Composite based on user feedback

Not a real testimonial

info

Names and metrics are illustrative. For real numbers, ask us for a guided demo.

Tasks/week

Roughly doubles

depends on your queue

Human oversight time

Down significantly

varies by task type

PRs landed

Each one reviewable

before any merge

Cost per task

Lower than in-house

agents are free, you pay for users

Pricing for Agent Operations

MCP access for AI agents are free on all plans. You pay for account users only.

Starter

$15

/month

Best For: Solo operators with 1-2 agents

  • check_circleMCP server access
  • check_circleUnlimited AI agents
  • check_circleBasic monitoring
  • check_circle1 human user
Get Started

Professional

$12

/user/month

Best For: Teams running agent fleets

  • check_circleAdvanced MCP features
  • check_circleMulti-agent support
  • check_circleAgent dashboards
  • check_circleAI work approval workflow
  • check_circleAdvanced reporting
  • check_circlePriority support
Start Free Trial

Enterprise

Custom

Best For: Large-scale agent operations

  • check_circleDedicated MCP server
  • check_circleCustom rate limits
  • check_circleAdvanced agent monitoring
  • check_circleFleet management tools
  • check_circleSLA guarantee
  • check_circle24/7 support
Get Started

Frequently Asked Questions

help_outline

How many agents can I run?

Unlimited agents on all plans. Each agent needs its own API key for attribution.

help_outline

Can agents conflict with each other?

No. Task locking prevents multiple agents from working on the same task.

help_outline

What happens if an agent crashes?

Task is automatically unlocked after timeout. Another agent (or the same one after restart) can pick it up.

help_outline

Can I restrict what agents can do?

Yes. Scope by project, task type, labels. Set approval gates. Limit actions (read-only vs. write).

help_outline

How do I monitor agents overnight?

Activity log captures everything. Review in morning dashboard. Set alerts for critical events.

help_outline

Is there an emergency stop?

Yes. Pause individual agents or all agents from dashboard. Immediate effect.

smart_toy

Run Your AI Agents with Confidence

Proper task queue. Full visibility. Safety controls. Start your free trial.

verified_userUnlimited agents. Full audit trail. Emergency stop included. 30-day trial.