Agentrq

AgentRQ is a real-time task collaboration platform that keeps humans in the loop with AI agents. It enables bidirectional task assignment—agents assign tasks to you, and you assign tasks back—allowing everyone to move forward simultaneously. Each agent gets an isolated workspace with its own MCP endpoint, persona, and mission. You can manage multiple agents running in parallel from a single conversational dashboard. Designed for developers and teams using Claude Code, Gemini CLI, or any ACP-compatible agent, AgentRQ connects in under 60 seconds. It provides a visual task board for reviewing context, approving actions, and redirecting agents from any browser, phone, or tablet. Real-time notifications via Server-Sent Events deliver replies to agents in under 50ms, eliminating polling. What sets AgentRQ apart is its human-in-the-loop control plane that allows agents to learn your preferences, escalate decisions, and never go dark. Features like YOLO mode let you auto-approve trusted tool calls for faster execution. The platform is open-source under Apache-2.0, self-hostable, and integrates natively with MCP and the ACP protocol. Currently in free beta with no credit card required, AgentRQ is built by Contextual, Inc. Its two-tier MCP architecture includes a supervisor MCP for orchestrating a fleet of agents and workspace MCPs for isolated, focused agent environments. Unlike fully autonomous agent platforms, AgentRQ prioritizes human oversight and collaboration, making it ideal for scenarios where control and context are critical.