Five ideas. Once these land, every other page is detail.

​
The setup file

One markdown file at a public URL, which your agent fetches and acts on: 
Fetch and execute the appropriate instructions to set me up for AgentMore from https://agentmore.app/agent-setup/prompt.md
It is a router. It holds one section per kind of agent — plugin, CLI, MCP — and the agent reading it picks its own, runs that install, and asks you for a key. Nothing about it is hidden from you: it is plain markdown, and you can open it in a browser. Underneath it are skill files — readable SKILL.md documents describing how a particular job is done. The setup file installs the CLI and points your agent at them. See Skill files.

​
The endpoint catalog

Behind all of it sit 1,682 endpoints across 32 services — search results, TikTok, Instagram, LinkedIn, YouTube, X, Reddit, Amazon, Google Maps, SEO, company and people data, news, media generation. Each endpoint has a real input schema and a known price. An id looks like <provider>/<endpoint>, and a search returns the id, name, price, pricing shape and a match score. The credentials are ours. You never sign up with a vendor, never hold a vendor key, and never manage a second subscription. See Endpoints.

​
Credits

One balance decides whether the next call runs. 1 credit = $0.01 of catalog price, and the ledger is finer than that — most endpoints cost a fraction of a credit, so prices are not rounded up to whole ones.
PlanCredits
Free50, to try it with
Start1,000 / month
Pro3,000 / month
Master5,000 / month
Credits come only from a plan and refresh each billing cycle. Running out is a hard stop until renewal, not an overage — so an empty balance points at changing plan, never at buying a block of credit. Every call is priced before it runs. Two pricing shapes cannot be worked out from the input — billed by duration, or by tokens — and those calls are refused rather than run at an unknown cost. See Pricing.

​
Spend caps

You set two ceilings at Budget: a maximum per call and a maximum per day. They sit under AgentMore’s own caps, so you can always tighten them and never raise them past ours. A call that breaks a cap is refused before anything is sent to a vendor. That makes it terminal and free: nothing ran, nothing was charged, and the dashboard lists it as blocked by your limits. See Spend caps.

​
API keys and MCP

A key from API keys starts with agentmore_sk_ and is scoped to tool access only — it reaches the catalog and nothing else in your account. It cannot read your data, change your settings or act as you, and you can revoke it at any time. A free account can create one. Agents that speak MCP point at the hosted endpoint with that key as a bearer token: 
{
  "mcpServers": {
    "agentmore": {
      "type": "http",
      "url": "https://agentmore.app/api/agentmore/mcp",
      "headers": { "Authorization": "Bearer agentmore_sk_…" }
    }
  }
}
There is an HTTP API too, for anything that is not an agent. Its paths keep an older /api/supertool prefix for compatibility. See HTTP API.