The agent library

Ready-made agents that run a whole function — QA, SEO, social and CRM — each with its own screens rather than just an answer.

How this differs from chat and workflows

Monokrew gives you three ways to put an agent to work, and they are not variations of the same thing.

The first two hand you an outcome. A library agent hands you somewhere to work.

That is the real difference. A social workflow can write and publish a post. The social agent gives you a posting calendar, an approval queue, a budget you set and watch, and performance across each platform — because running social media is not one job, it is a function with its own shape.

Each library agent is a part of a company, managed by agents, with the screens that part of a company needs.

What that means in practice

• Its own interface. Built for that function, not a generic run-and-result view. A calendar where a calendar belongs; a report where a report belongs.
• Its own actions. Things you approve, adjust, pause or override — inside the surface, while the agent works.
• Its own measures. Each agent shows whether it is working: what ran, what it produced, what it cost, and what changed as a result.

Turning one on

Enable the agent you want, then make sure it has what it needs — the skills, MCP servers and toolkits that function depends on. An agent that cannot reach your accounts cannot do the job.

The agent tells you what is missing rather than working around it. See Secrets, MCP servers and toolkits for where each connection lives.

Tip — Enable one agent and let it run for a few weeks before adding another. Each one takes a little setup and a little attention at the start, and the first one teaches you how the rest behave.

The agents

The first four cover testing, search, social and customers. Each has its own page.

• QA — documentation coming as the surface ships
• SEO — documentation coming as the surface ships
• Social (SMM) — documentation coming as the surface ships
• CRM — documentation coming as the surface ships

More agents follow. Each one arrives with its own page here.

Warning — These pages are written when each agent's screens are finished, not before. A page describing a screen that does not exist yet is worse than no page: it sends you looking for a button that is not there.

What it costs

Library agents spend billing tokens for the work they do, like anything else. An agent running continuously — testing on a schedule, auditing weekly, managing a calendar — spends more than an occasional request, and what it spends shows on its own screens.

See How billing tokens work.

Next steps

To connect what an agent needs before enabling it, continue to Secrets, MCP servers and toolkits.