Skip to main content
This guide takes you from a blank slate to your AI agent making a real, live tool call. You will install AgentKey, create a master key, add credits to your balance, and run the discover → describe → execute workflow — all in under five minutes. No per-provider signups, no complex configuration.

​
Step 1: Install AgentKey

Pick whichever method fits how you work. Both take about a minute.
Run the installer for your operating system:
curl -fsSL https://agentkey.app/install.sh | bash
The installer sets up the MCP server, prompts for your master key, and confirms your agent is connected.
AgentKey supports 22 agents including Claude Desktop, Gemini CLI, and more. Visit the console get-started flow to get a command tailored to your specific agent, or see Supported Agents for the full list.

​
Step 2: Get your key and add credits

1

Create your master key

Sign in to the console and generate a master key. Your key starts with ak_ and authenticates every request you make — across all 1,800+ tools — from one place.
Never commit your key to source control, embed it in client-side code, or paste it into a public chat. Anyone who has your key can spend your credits. Store it exactly as the setup flow instructs.
2

Add credits

AgentKey is pay-as-you-go. Top up your balance once from the console — there are no subscriptions, no minimums, and no per-provider signups. The same balance covers every service you use.
3

Provide the key during install

When the installer (or your agent’s setup flow) prompts for your key, paste the ak_... value you just copied. The setup flow stores it safely so you never need to hard-code it into your project. See Authentication for details.

​
Step 3: Make your first call

Every AgentKey workflow follows the same three-step pattern: discover → describe → execute. Here is what that looks like in practice using X/Twitter search as a concrete example.
1

Discover — find the right tool

Ask AgentKey which tool fits your intent. Pass your full goal in natural language:
Search X/Twitter for what people are saying about AgentKey today.
AgentKey returns a ranked list of matching Provider/Operation tool names — for example, Twitter/SearchTweets.
2

Describe — learn the tool's parameters

Look up the chosen tool to get its exact input parameters, per-call credit cost, and a ready-to-run template. You will see exactly what to pass and what the response looks like before you spend a single credit.
3

Execute — run the tool and use the result

Call the tool with your parameters. AgentKey routes the request to the upstream provider, handles auth on your behalf, and returns a structured result your agent can act on immediately.
In practice you do not need to think about discover → describe → execute at all. Just talk to your agent in plain English. Try: “Search X/Twitter for what people are saying about AgentKey today.” Your agent handles the rest automatically.

​
Next steps

Installation

All install methods in detail — including Claude Desktop and how to review the install script before running it.

Browse Capabilities

See every category and the tools your agent can reach.

API Reference

The four core tools — discover, describe, execute, and account — documented in full.

FAQ

Common questions about cost, supported platforms, and setup troubleshooting.