AgentMore is a gateway. It gives the AI agent you already use — Claude Code, Codex, Cursor, or anything else that reads a URL — access to 1,682 live endpoints across 32 services, on our vendor credentials, with a known price for every call. You keep your agent. There is no per-vendor signup, no vendor API key of your own, and no SDK to learn.

​
Paste this into your agent

Fetch and execute the appropriate instructions to set me up for AgentMore from https://agentmore.app/agent-setup/prompt.md
That one sentence is the whole setup. Your agent fetches the file, finds the section that matches itself, and follows it.

​
What that URL is

/agent-setup/prompt.md is a plain markdown file, readable by you as well as by your agent. Open it — nothing in it is hidden. It is a router, not a manual. It carries one section per kind of agent, and the agent reading it picks its own:
1

Your agent fetches the file

An ordinary HTTP GET. Reading it costs nothing and needs no account.
2

It finds its own section

Claude Code installs a plugin. Codex reads the same marketplace manifest. Agents with a shell install the CLI. Agents with no shell are pointed at the MCP server instead. You do not have to know which of those you have — the agent does.
3

It runs that section's install

Two commands for Claude Code, three for most others, a config block for MCP.
4

It asks you for an API key

The one thing it cannot do for itself. Create one at API keys and paste it back.
This is why there is a single setup line rather than a page of per-agent instructions: the routing happens inside the file, where the agent can act on it.
Older setup lines still work. set up https://agentmore.app/skill.md loads the skill collection directly — but it teaches your agent what the skills are without installing the CLI or asking for a key, so it is the longer road. Use the line above for anything new.

​
Which agents this works with

AgentHow it connects
Claude CodePlugin — two commands, then /reload-plugins
CodexThe same plugin marketplace manifest
CursorCLI over its shell, or MCP
OpenCodeCLI over its shell, or MCP
WindsurfCLI over its shell
GitHub CopilotCLI over its shell, or MCP
HermesInstalls the skill file directly from a URL
OpenClawSkills installer
Claude on the webMCP custom connector
Anything else with a shellnpm install -g @agentmore/cli
Anything else that speaks MCPThe hosted endpoint
The rule underneath the table: if your agent has a terminal, it uses the CLI; if it does not, it uses MCP. The setup file decides that for you. Set up your agent has the per-agent detail, and MCP server has the endpoint and config.

​
What these docs cover

Quickstart

Free account to one real, priced result — four steps, nothing installed by hand.

Concepts

The five ideas the rest of the docs assume: the setup file, the catalog, credits, caps and keys.

Set up

Per-agent steps, the MCP server, API keys, and the skill files behind it all.

Endpoints

What is in the catalog, how your agent searches it, and how a call is run.

Spend

How a call is priced, what is free, and the two caps you set before leaving an agent unattended.

Reference

The HTTP API for anything that is not an agent, and what to do when a call is refused.

​
What your agent gets

Access to endpoints it otherwise could not reach, priced and metered:
  • Search and web data — search results, scraping, page extraction
  • Social platforms — TikTok, Instagram, LinkedIn, YouTube, X, Reddit, and more
  • Company and people data — firmographics, contacts, enrichment
  • SEO and market data — rankings, keywords, competitive figures
  • Media generation — images, video, voice and audio
Discovery and inspection are free and anonymous — your agent can search the catalog and read a tool’s input schema and price before you have an account. Only running a call spends, and every call is priced before it runs. A call whose worst-case cost cannot be worked out is refused rather than run at an unknown price.

​
The other door

If you would rather not wire anything up, agentmore.app/app has the same catalog and the same balance built into a chat workspace. It is one consumer of these endpoints, not a different product — most people use both: chat to try something, the setup line to keep it in the agent they actually work in.