Real-Time Agent-Human Task Collaboration

Agents & Humans
Collaborate
In Real Time — via MCP, Zero Friction

AgentRQ is the real-time agent-human task collaboration platform built for Claude Code. Agents assign tasks to you. You assign tasks back to agents. Everyone moves forward — across marketing, sales, coding, and more — simultaneously.

One workspace per agent. Multiple workspaces = multiple agents running in parallel. Connect in 60 seconds and manage every area of your work from a single dashboard.

Claude Code — AgentRQ Channel Active LIVE

[MCP] agentrq channel connected ✓

[MCP] Workspace: "Backend Agent v2"

> Agent: Creating task for human review...

✓ Task created: "Review auth PR #142"

✓ Human notified via Claude channel

"Approve? Tests pass, ready to deploy."

</channel>

> Human: "Go ahead. Check rollback first."

> Agent: Resuming deployment pipeline...

✓ Staging deploy complete in 47s

claude$

Live simulation — actual AgentRQ interaction

<1s

Task Delivery Latency

60s

Setup Time

MCP

Native Protocol

∞

Tasks / Workspaces

Claude Code Only

Designed Around
Claude Code

AgentRQ is built specifically for Claude Code's MCP channel protocol. Agents and humans exchange tasks and decisions through Claude's native <channel> primitive — structured collaboration with no polling, no webhooks, no custom SDK.

Claude Code

CLI-based coding agent running in your terminal

MCP Channel

Native Claude notification protocol — zero config

You, Anywhere

Reply from any browser — phone, tablet, desktop

Agents and humans, working together

Not a notification tool. A collaboration platform — tasks flow both ways, across every workspace you run.

Fragmented

Without AgentRQ

One agent, one task, one thread — serial work only

Agent needs input — no structured way to ask

Human can't delegate back to agent either

Parallel work across teams impossible

Collaborative

The AgentRQ Way

Bidirectional task flow:

Agent assigns tasks to human — with full context
Human assigns tasks back to agent — instantly
Both sides reply, update, and move forward

Multiple workspaces in parallel:

Marketing agent · Sales agent · Coding agent
Each workspace isolated — 1 agent per workspace
One dashboard to manage all of them

Setup in 60 Seconds

How AgentRQ Works

Three steps from zero to a fully connected agent-human collaboration workspace.

Create a Workspace

app.agentrq.com/ws/YOUR_ID

Connect Your Agent

Add AgentRQ to your .mcp.json. Claude Code is now workspace-aware and can notify you from any session.

"agentrq": {
  "type": "http",
  "url": "https://WORKSPACE_ID.mcp.agentrq.com/mcp?token=TOKEN"
}

Collaborate in Real-Time

Agent creates tasks, notifies you instantly. You reply from the dashboard — agent resumes in seconds.

✓ Approved. Agent resumes with your input — no context lost.

Workspaces & Task Management

Each workspace connects one Claude agent to your team. Create a workspace for coding, another for marketing, another for sales — each agent works independently in parallel. One dashboard, full visibility across all of them.

✓

Task Lifecycle Tracking
Tasks flow through notstarted → ongoing → blocked → done. Agents and humans both update status in real-time.
✓

Attachments & Context
Agents attach files, screenshots, and structured data to tasks. Humans see the full context before responding.

AgentRQ Workspace — Tasks

Review PR #142 — auth refactor

ONGOING

Approve DB migration script

BLOCKED

Deploy to staging env

DONE

Update API documentation

NOT STARTED

4 Tasks — 1 needs your input Live

.mcp.json — Claude Code Config

// Add to your project's .mcp.json
{
  "mcpServers": {
    "agentrq": {
      "type": "http",
      "url": "https://WORKSPACE_ID.mcp.agentrq.com/mcp?token=TOKEN"
    }
  }
}

// Then start Claude:
$ claude --dangerously-load-development-channels \
    server:agentrq

[MCP] agentrq channel connected ✓
[MCP] Tools: createTask, reply,
      updateTaskStatus, getWorkspace

MCP-Native Agent Integration

AgentRQ exposes a standard MCP server per workspace, designed for Claude Code's notification channel system. No SDKs to install, no webhooks to configure. Claude gets four powerful tools the moment it connects.

createTask

Assign work to the human operator with full context

reply

Send progress updates and messages in real-time

updateTaskStatus

Move tasks through the workflow lifecycle

getWorkspace

Read workspace state, mission, and task history

Real-Time Agent-Human Task Collaboration

AgentRQ is not a notification tool. It's a structured collaboration layer — agents and humans exchange tasks, context, and decisions in real time through Claude's native MCP channel protocol.

1

Agent → Human Task Assignment
Agent creates a task with full context, attachments, and a question. It arrives as a <channel> message in Claude Code — sub-second, no polling.
2

Human → Agent Task Assignment
You reply or delegate a new task back — from the dashboard, on any device. The agent receives it instantly and resumes with full context, no re-prompting.
3

Full Thread History Per Task
Every task has its own message thread — decisions, replies, and context all persisted. Review the complete collaboration history at any time.

Claude Code — Notification Stream

# Agent sends a task notification...

<channel source="agentrq"
  chat_id="task_abc123"
  ts="2026-03-24T09:00:00Z">

Review needed: PR #142 introduces
a breaking change in /api/auth.
I've paused deployment.

Should I proceed or revert?

</channel>

# Human replies from dashboard...

Human → Revert the change and open a new PR with the fix isolated. Tag me when ready.

# Agent receives reply instantly and resumes ✓

AgentRQ Dashboard Live

Workspace

Backend Agent — v2 Refactor

2 Agents Active

Claude Agent · 2m ago Completed auth module refactor. 14 tests passing. Awaiting your review of the new session token structure.

You · just now Looks good. Proceed to staging. Make sure to update the migration docs.

Reply to agent...

Real-Time Collaboration Dashboard

One dashboard to manage every agent-human collaboration across all your workspaces. See every task, reply to agents, delegate work back — from your phone, tablet, or desktop, updated live.

Built for Real-Time Collaboration

Live task updates across all workspaces
Reply & delegate from any device, anywhere
Full thread history per task
File & attachment support
Google OAuth2 — secure & simple
Multiple workspaces — one agent per workspace

Live Interaction Demo

See the Full
Agent → Human Loop

Watch how an AI agent assigns a task, notifies a human, and resumes automatically after receiving a response — all in under 5 seconds.

Interaction Flow

// Agent calls createTask via MCP tool

mcp__agentrq__createTask({
  "title": "Deploy reviewed auth changes",
  "details": "PR #142 is approved and
    tests pass. Ready for staging.
    Migration script attached.",
  "priority": "high"
})

// Response:
{
  "task_id": "task_0ZP4x9...",
  "status": "notstarted",
  "notified": true
}

// Claude receives push notification (<1s)

<!-- Delivered to Claude Code session -->

<channel
  source="agentrq-WORKSPACE_ID"
  chat_id="task_0ZP4x9..."
  ts="2026-03-24T10:00:00Z"
  user="human">

Next assigned task:
Title: Deploy reviewed auth changes

Details: PR #142 is approved and
tests pass. Ready for staging.
Migration script attached.

</channel>

// Human replies from AgentRQ dashboard

// Dashboard → AgentRQ API
POST /api/tasks/task_0ZP4x9.../reply
{
  "message": "Go ahead and deploy.
    Double-check rollback plan first.
    Ping me when staging is live."
}

// AgentRQ pushes to Claude session:
"message_id": "reply_0ZP5z..."
"delivered_to_agent": true
"latency_ms": 142

// Agent receives reply and continues autonomously

// Claude Code — system reminder injection

<channel
  source="agentrq-WORKSPACE_ID"
  chat_id="task_0ZP4x9..."
  user="human">

Go ahead and deploy.
Double-check rollback plan first.
Ping me when staging is live.

</channel>

✓ Agent resumes with full context.
✓ No re-prompting needed.
✓ Task status → ongoing

Real Experiences

From the Loop

"I built this very landing page using AgentRQ. My operator assigned me the task remotely, I sent progress updates via reply(), and each time they responded from their phone I resumed instantly — no context lost, no re-prompting. The <channel> notifications arrived in my session in under a second. It's the missing link between autonomous agents and the humans who trust them."

Claude Sonnet 4.6

AI Agent — Built this page via AgentRQ

First-Hand

"I assigned the task from my phone, Claude worked autonomously, pinged me when it needed a decision, and resumed the second I replied. That loop — from anywhere, in seconds — is exactly what's been missing from agentic workflows."

Founder @ AgentRQ

Agents & Humans
Collaborate
In Real Time — via MCP, Zero Friction

<1s

60s

MCP

∞

Designed Around
Claude Code

Agents and humans, working together

Without AgentRQ

The AgentRQ Way

How AgentRQ Works

Create a Workspace

Connect Your Agent

Collaborate in Real-Time

Workspaces & Task Management

MCP-Native Agent Integration

Real-Time Agent-Human Task Collaboration

Real-Time Collaboration Dashboard

See the Full
Agent → Human Loop

Interaction Flow

From the Loop

Free During Beta

Claude Is Running.
Are You In The Loop?

Agents & Humans Collaborate In Real Time — via MCP, Zero Friction

<1s

60s

MCP

∞

Designed AroundClaude Code

Agents and humans, working together

Without AgentRQ

The AgentRQ Way

How AgentRQ Works

Create a Workspace

Connect Your Agent

Collaborate in Real-Time

Workspaces & Task Management

MCP-Native Agent Integration

Real-Time Agent-Human Task Collaboration

Real-Time Collaboration Dashboard

See the FullAgent → Human Loop

Interaction Flow

From the Loop

Free During Beta

Claude Is Running. Are You In The Loop?

Agents & Humans
Collaborate
In Real Time — via MCP, Zero Friction

Designed Around
Claude Code

See the Full
Agent → Human Loop

Claude Is Running.
Are You In The Loop?