Overlord -- AI Operations Commander

A single agent is a demo. A fleet is an operations problem.

The moment you run more than one agent, the hard parts stop being prompts. They become the things nobody writes tutorials about: which agent owns what, where state lives between them, what happens when one dies mid-task, and why last night's run cost four times last week's.

Overlord is the operations layer for that. Four skills that assume the fleet already exists and is already causing you trouble.

What's In It

• Phantom -- Autonomous Agent Builder. Standing up agents that recover on their own instead of stalling silently at 3am.
• Hivemind -- Multi-Agent Orchestrator. Dispatch topology, work handoff, and the coordination rules that keep two agents from claiming the same job.
• Recall -- Agent Memory Engineer. Persistent memory that survives restarts, with expiry policy so context does not grow until it collapses.
• Anthropic -- Claude API Integration. Model routing, token budgeting, and prompt caching -- the difference between a fleet that costs $40 a month and one that costs $400.

The math

| Bought separately | | |---|---| | Phantom | $19 | | Hivemind | $19 | | Recall | $19 | | Anthropic | $29 | | Total | $86 | | Overlord bundle | $69 | | You save | $17 (20%) |

Who this is for

You run two or more agents in something resembling production. You have already been surprised by a bill, a stalled overnight run, or two agents fighting over the same file.

Who this is not for

You are running one agent and it works. Nothing here will help yet -- buy Phantom on its own when you add the second.

Set the architecture first with Architect, then run it with Overlord.

Works with OpenClaw and any agent framework.