Skip to main content
AgentKey exposes a small, stable set of tools rather than ~1,800 individual endpoints. You discover the underlying tool you need, describe it to learn its parameters and cost, then execute it.

find_tools

Discover tools by natural-language intent.

describe_tool

Get a tool’s parameters, cost, and run template.

execute_tool

Run a tool by its Provider/Operation name.

account

Check credits and upstream health — free.

​
Authentication

All calls authenticate with your master key. You don’t pass it by hand on each call — the installer or skill setup configures it once. See Authentication.

​
The standard workflow

1

Discover

find_tools (semantic) or list_tools (browse) → returns Provider/Operation names.
2

Describe

describe_tool → params, per-call cost, and an execute_as template.
3

Execute

execute_tool → runs the tool and returns the result.
Make one execute_tool call per turn and await the result before the next — don’t batch. Never guess a tool name or its parameters; always describe first.
This reference describes the tool contracts as exposed through the MCP server. Exact REST paths and request envelopes are confirmed in the console.