Human-in-the-Loop Task Manager for AI Agents

Guidelines for Running AI Coding Agents

Short, practical guides for running AI coding agents well. Most guidelines cover one setting across many agents, with the exact flag, config key or toggle for each one, checked against that agent's own documentation or source code. Others cover connecting agents to AgentRQ and keeping your workspaces secure.

The guidelines start with YOLO mode, the setting that lets an agent act without asking for approval. Its guides cover eleven agents, from Codex CLI and Claude Code to Antigravity, Gemini CLI, GitHub Copilot, Cline and OpenCode, and each one ends with how to approve that agent's requests remotely, per task, from AgentRQ.

The connect agents guidelines show, click by click, how to connect Claude Code with machines and how to connect an agent with the ACP Gateway and machines: enrol a computer once and start Claude Code, Antigravity, Gemini CLI or any ACP agent from AgentRQ, or run it yourself. If a running Claude Code drops off and its workspace goes offline, reconnect its AgentRQ MCP from the terminal. On a Claude Team or Enterprise plan, an admin first has to enable Claude Channels.

The task manager guidelines explain what a task manager for AI agents needs and why: Task Manager for AI Agents and Characteristics covers memory, skills, workflows, schedulers, tools with Mesher and human controls, what each one unlocks, and a checklist for both agents and people.

The task guidelines show every way to hand work to an agent: create a task in a workspace, schedule a one-time task, schedule a repeating task, and create tasks from a Pano doc.

The team guidelines show how a team works together on AgentRQ: see what your teammates' agents are working on, share agent memory across a team so each agent reads what the others learned, use a teammate's skills, and create a team and add a new member.

The security guidelines cover what to do when an agent credential leaks, starting with a compromised .mcp.json file: reset the workspace ID to revoke its token and keep your tasks, or purge the workspace.

The workspace guidelines explain how to read a workspace at a glance, starting with what each workspace status color means: which dot is waiting on you, which is working, which is stuck, and what a pulsing or rippling dot tells you. They also walk through every tab of a workspace's settings, from General and Setup to Skills, Sharing and the Danger Zone.

Start Free