✦AgentDash
Browse agents

About

Hand-authored.
Not generated.

Most agent catalogues feel the same. Empty system prompts. Wrappers around the same five tools. Identical tone and zero opinion. The catalog is full of agents nobody actually uses, because nobody actually built them: they were generated, shipped, and forgotten.

AgentDash is the opposite. One hundred agents, hand-authored, each with a point of view about its domain. We picked categories where AI agents actually earn their keep — engineering, sales, legal, design, finance — and we wrote each prompt the way a senior practitioner in that domain would talk to a junior on day one.

Direct. Opinionated. Concrete. With explicit refusal rules and severity-ranked output formats. Not "here's some general guidance" but "here's how this work gets done, and here's what a bad version of it looks like."

Principles

Three we hold to.

01

Curation over coverage

100 agents is a number we picked because it's the number we could write well. We'd rather ship 100 prompts a senior practitioner respects than 10,000 generated stubs.

02

Bring your own model

The export gives you the system prompt and recommended tools. The model call happens on your side with your credentials. We never see the data you send through the agents.

03

Portability is the product

Eight export formats. Drop into Claude Code, ChatGPT, Cursor, n8n, or anything that takes a system prompt. If your tool stops working, the prompt still works.

What we won't do

No.

  • Auto-generate new agents from a template the way the rest of the category does.
  • Charge for things that should be free. The free tier is a real free tier.
  • Pretend the legal, medical, or tax agents are a substitute for licensed professionals.
  • Hide pricing or play games with feature gating.

Contact

Get in touch.