Skip to main content
This guide takes you from nothing to your agent’s first live tool call.

​
1. Install AgentKey

Pick whichever fits how you work — each takes about a minute.
Run the installer for your OS. It detects the agents on your machine, installs the AgentKey skill into each one, and opens a browser tab to sign in — no key to copy, no JSON to edit:

Full installation guide

All install methods in detail — including Claude Desktop — plus how to review the install script first.
AgentKey works with many agents. The console get-started flow matches your agent and gives you the exact command; see Supported Agents for the full list.

​
2. Get your key and pick a plan

1

Create a key

Sign in to the console and create a master key (ak_...). Treat it like a password — it authenticates every request and is tied to your subscription.
Never commit your key to source control or paste it into a chat. Store it as instructed by the setup flow.
2

Pick a plan

AgentKey is subscription-based. Choose a plan from the console — each plan includes a monthly credit allowance shared across every service; add extra credits or upgrade when you need more. No per-provider signups. Plans and prices are on the pricing page.
3

Sign in

The installer (and the npx path) opens a browser tab to sign in to AgentKey — that is the authentication. Only clients that cannot do MCP OAuth need an API key pasted into their config. See Authentication.

How billing works

Learn how plans, monthly credits, and per-call costs work.

​
3. Make your first call

Every AgentKey workflow follows the same pattern: discover → describe → execute.
1

Discover

Ask AgentKey which tool fits your task. Pass your full intent in natural language — e.g. “search the web for the latest Claude news”. You get back ranked Provider/Operation tool names.
2

Describe

Look up the chosen tool to get its exact parameters, per-call cost, and a ready-to-run template.
3

Execute

Run the tool with your parameters and use the result.
In practice you just ask your agent in plain language. Try: “Search X/Twitter for what people are saying about AgentKey today.”
Calling AgentKey from code instead of an agent? Skip the install: the Scenario Methods REST API takes your key as a bearer token and returns fixed-schema results.

Understand the model

Why AgentKey uses discover → describe → execute, and how to use it well.

​
Next steps

Installation

The two install methods in detail.

Browse capabilities

See everything your agent can reach.

API reference

Scenario Methods over REST, and the four MCP tools.

FAQ

Common questions about cost, platforms, and setup.