AgentRQ

AgentRQ is a human-in-the-loop task manager for AI agents. It is built for Claude Code and ACP-compatible agents, including Gemini, and is designed for task collaboration in which agents assign work to humans and humans assign work back to agents. The product describes itself as a real-time control plane for keeping people involved while agents work autonomously.

Its core features center on task coordination and context management. AgentRQ says agents can learn user preferences, escalate decisions, and never go dark. It provides one workspace per agent, with multiple workspaces used to run multiple agents in parallel. The interface includes a conversational control panel where tasks can be created, replies can be sent to agents, and threads and decisions remain available in one place. The page also mentions a task board with statuses such as ongoing, blocked, and not started, as well as real-time notifications and a YOLO mode listed among the features.

AgentRQ also exposes MCP-based coordination. The service says a single MCP server at mcp.agentrq.com can be used to assign tasks to different agents, and that each workspace gets a dedicated MCP endpoint. It states that the product works with Gemini CLI and all ACP-compatible agents, supports acp-gateway, provides native push notifications for Claude Code users, and connects OpenAI Codex agents through a Codex app gateway. The site repeatedly frames the product around Claude, Gemini, and other ACP agents, with a focus on supervision and parallel management from a single dashboard.

The service is free during beta and says no credit card is required to start. It is described as open source, Apache-2.0 licensed, and self-hostable. The page also identifies Contextual, Inc. in the footer.